In [ ]:
import numpy as np
import pandas as pd
from packaging import version
import time

from sklearn.metrics import confusion_matrix, classification_report
from sklearn.metrics import accuracy_score
from sklearn.metrics import mean_squared_error as MSE
from sklearn.model_selection import train_test_split

import matplotlib.pyplot as plt
import seaborn as sns

import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import models, layers
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D, MaxPool2D, BatchNormalization, Dropout, Flatten, Dense
from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping
from tensorflow.keras.preprocessing import image
from tensorflow.keras.utils import to_categorical
import tensorflow.keras.backend as k

%matplotlib inline
np.set_printoptions(precision=3, suppress=True)

print("This notebook requires TensorFlow 2.0 or above")
print("TensorFlow version: ", tf.__version__)
assert version.parse(tf.__version__).release[0] >=2

print("Keras version: ", keras.__version__)

This notebook requires TensorFlow 2.0 or above
TensorFlow version:  2.18.0
Keras version:  3.8.0
In [ ]:
# Loading the cifar10 Dataset
(train_images, train_labels), (test_images, test_labels) = keras.datasets.cifar10.load_data()

Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz
170498071/170498071 ━━━━━━━━━━━━━━━━━━━━ 15s 0us/step
In [ ]:
# Exploratory Data Analysis
print(f"train images shape`: {train_images.shape}")
print(f"train labels shape: {train_labels.shape}")
print(f"test images shape: {test_images.shape}")
print(f"test labels shape: {test_labels.shape}")

train images shape`: (50000, 32, 32, 3)
train labels shape: (50000, 1)
test images shape: (10000, 32, 32, 3)
test labels shape: (10000, 1)
In [ ]:
# Explore the labels, labeled as a numerical digit that needs conversion
# to an item description
print(f"First 10 training labels: {train_labels[:10]}")

First 10 training labels: [[6]
 [9]
 [9]
 [4]
 [1]
 [1]
 [2]
 [7]
 [8]
 [3]]
In [ ]:
# Data Analysis Functions

def show_random_examples(x, y, p):
    indices = np.random.choice(range(x.shape[0]), 10, replace=False)

    x = x[indices]
    y = y[indices]
    p = p[indices]

    plt.figure(figsize=(10, 5))
    for i in range(10):
        plt.subplot(2, 5, i + 1)
        plt.imshow(x[i])
        plt.xticks([])
        plt.yticks([])
        col = 'green' if np.argmax(y[i]) == np.argmax(p[i]) else 'red'
        plt.xlabel(class_names_preview[np.argmax(p[i])], color=col)
    plt.show()

def get_three_classes(x, y):
    def indices_of(class_id):
        indices, _ = np.where(y == float(class_id))
        return indices

    indices = np.concatenate([indices_of(0), indices_of(1), indices_of(2)], axis=0)

    x = x[indices]
    y = y[indices]

    count = x.shape[0]
    indices = np.random.choice(range(count), count, replace=False)

    x = x[indices]
    y = y[indices]

    y = tf.keras.utils.to_categorical(y)

    return x, y

def plot_history(history):
  losses = history.history['loss']
  accs = history.history['accuracy']
  val_losses = history.history['val_loss']
  val_accs = history.history['val_accuracy']
  epochs = len(losses)

  plt.figure(figsize=(16, 4))
  for i, metrics in enumerate(zip([losses, accs], [val_losses, val_accs], ['Loss', 'Accuracy'])):
    plt.subplot(1, 2, i + 1)
    plt.plot(range(epochs), metrics[0], label='Training {}'.format(metrics[2]))
    plt.plot(range(epochs), metrics[1], label='Validation {}'.format(metrics[2]))
    plt.legend()
  plt.show()

def display_training_curves(training, validation, title, subplot):
  ax = plt.subplot(subplot)
  ax.plot(training)
  ax.plot(validation)
  ax.set_title('model '+ title)
  ax.set_ylabel(title)
  ax.set_xlabel('epoch')
  ax.legend(['training', 'validation'])

def print_validation_report(y_test, predictions):
    print("Classification Report")
    print(classification_report(y_test, predictions))
    print('Accuracy Score: {}'.format(accuracy_score(y_test, predictions)))
    print('Root Mean Square Error: {}'.format(np.sqrt(MSE(y_test, predictions))))


def plot_confusion_matrix(y_true, y_pred):
    mtx = confusion_matrix(y_true, y_pred)
    fig, ax = plt.subplots(figsize=(16,12))
    sns.heatmap(mtx, annot=True, fmt='d', linewidths=.75,  cbar=False, ax=ax,cmap='Blues',linecolor='white')
    #  square=True,
    plt.ylabel('true label')
    plt.xlabel('predicted label')
In [ ]:
train_image_preview, train_label_preview = get_three_classes(train_images, train_labels)
test_image_preview, test_label_preview = get_three_classes(test_images, test_labels)

class_names_preview = ['airplane', 'car', 'bird']
show_random_examples(train_image_preview, train_label_preview, train_label_preview)
No description has been provided for this image
In [ ]:
class_names = ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog','frog', 'horse' ,'ship' ,'truck']
In [ ]:
image_train_split, image_val_split, label_train_split, label_val_split = train_test_split(train_images, train_labels, test_size=.1, random_state=42, shuffle=True)
print(image_train_split.shape)
print(image_val_split.shape)
print(label_train_split.shape)
print(label_val_split.shape)

(45000, 32, 32, 3)
(5000, 32, 32, 3)
(45000, 1)
(5000, 1)
In [ ]:
image_train_norm = image_train_split / 255.0
image_val_norm = image_val_split / 255.0
image_test_norm = test_images / 255.0
image_train_norm.shape
Out[ ]:
(45000, 32, 32, 3)
In [ ]:
# Build Model 2-1: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_8'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=8, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 8s 6ms/step - accuracy: 0.3254 - loss: 1.8834 - val_accuracy: 0.4512 - val_loss: 1.5801
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 3ms/step - accuracy: 0.4740 - loss: 1.4991 - val_accuracy: 0.4938 - val_loss: 1.4709
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 2s 3ms/step - accuracy: 0.5024 - loss: 1.4143 - val_accuracy: 0.5060 - val_loss: 1.4343
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.5247 - loss: 1.3617 - val_accuracy: 0.5222 - val_loss: 1.3893
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.5331 - loss: 1.3317 - val_accuracy: 0.5412 - val_loss: 1.3437
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5447 - loss: 1.3119 - val_accuracy: 0.5492 - val_loss: 1.3248
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 2s 3ms/step - accuracy: 0.5478 - loss: 1.2959 - val_accuracy: 0.5436 - val_loss: 1.3217
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 2s 3ms/step - accuracy: 0.5500 - loss: 1.2830 - val_accuracy: 0.5588 - val_loss: 1.2985
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.5569 - loss: 1.2697 - val_accuracy: 0.5638 - val_loss: 1.2977
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.5542 - loss: 1.2741 - val_accuracy: 0.5518 - val_loss: 1.2975
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5574 - loss: 1.2619 - val_accuracy: 0.5696 - val_loss: 1.2681
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 3ms/step - accuracy: 0.5700 - loss: 1.2378 - val_accuracy: 0.5688 - val_loss: 1.2598
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.5710 - loss: 1.2365 - val_accuracy: 0.5604 - val_loss: 1.2711
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 2s 3ms/step - accuracy: 0.5656 - loss: 1.2517 - val_accuracy: 0.5622 - val_loss: 1.2742
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.5736 - loss: 1.2652
313/313 ━━━━━━━━━━━━━━━━━━━━ 0s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.71      0.54      0.61      1000
           1       0.69      0.68      0.69      1000
           2       0.40      0.45      0.42      1000
           3       0.38      0.49      0.43      1000
           4       0.48      0.56      0.51      1000
           5       0.53      0.35      0.42      1000
           6       0.72      0.64      0.68      1000
           7       0.58      0.67      0.62      1000
           8       0.75      0.63      0.69      1000
           9       0.59      0.66      0.63      1000

    accuracy                           0.57     10000
   macro avg       0.58      0.57      0.57     10000
weighted avg       0.58      0.57      0.57     10000

Accuracy Score: 0.5668
Root Mean Square Error: 2.8229594400203486
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build initial data for table
data = {}

data['model'] = ['DNN']
data['accuracy'] = [f"{history.history['accuracy'][-1]:.3f}"]
data['val_accuracy'] = [f"{history.history['val_accuracy'][-1]:.3f}"]
data['test_accuracy'] = [f"{test_pred[1]:.3f}"]
data['loss'] = [f"{history.history['loss'][-1]:.3f}"]
data['val_loss'] = [f"{history.history['val_loss'][-1]:.3f}"]
data['test_loss'] = [f"{test_pred[0]:.3f}"]
data['time'] = [f"{time_end - time_start:.3f}"]
In [ ]:
def add_to_data(data, model, history, test_pred):
    data['model'].append(model)
    data['accuracy'].append(f"{history.history['accuracy'][-1]:.3f}")
    data['val_accuracy'].append(f"{history.history['val_accuracy'][-1]:.3f}")
    data['test_accuracy'].append(f"{test_pred[1]:.3f}")
    data['loss'].append(f"{history.history['loss'][-1]:.3f}")
    data['val_loss'].append(f"{history.history['val_loss'][-1]:.3f}")
    data['test_loss'].append(f"{test_pred[0]:.3f}")
    data['time'].append(f"{time_end - time_start:.3f}")
In [ ]:
# Build Model 2-1: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_16'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=16, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 11ms/step - accuracy: 0.3473 - loss: 1.8265 - val_accuracy: 0.5032 - val_loss: 1.4674
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.5132 - loss: 1.4004 - val_accuracy: 0.5448 - val_loss: 1.3570
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 2s 3ms/step - accuracy: 0.5513 - loss: 1.2912 - val_accuracy: 0.5572 - val_loss: 1.2955
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5714 - loss: 1.2333 - val_accuracy: 0.5794 - val_loss: 1.2424
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5843 - loss: 1.2055 - val_accuracy: 0.5846 - val_loss: 1.2308
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 3ms/step - accuracy: 0.5938 - loss: 1.1756 - val_accuracy: 0.5920 - val_loss: 1.1989
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 2s 3ms/step - accuracy: 0.6023 - loss: 1.1435 - val_accuracy: 0.5878 - val_loss: 1.2063
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6058 - loss: 1.1326 - val_accuracy: 0.5900 - val_loss: 1.1924
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6106 - loss: 1.1261 - val_accuracy: 0.6054 - val_loss: 1.1510
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 3ms/step - accuracy: 0.6189 - loss: 1.1019 - val_accuracy: 0.6076 - val_loss: 1.1627
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.6207 - loss: 1.0961 - val_accuracy: 0.6114 - val_loss: 1.1377
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6268 - loss: 1.0697 - val_accuracy: 0.6192 - val_loss: 1.1221
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 3ms/step - accuracy: 0.6307 - loss: 1.0698 - val_accuracy: 0.6228 - val_loss: 1.1136
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 2s 3ms/step - accuracy: 0.6310 - loss: 1.0649 - val_accuracy: 0.6112 - val_loss: 1.1232
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.6399 - loss: 1.0379 - val_accuracy: 0.6144 - val_loss: 1.1165
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6345 - loss: 1.0475 - val_accuracy: 0.6100 - val_loss: 1.1104
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.6214 - loss: 1.0937
313/313 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.59      0.74      0.66      1000
           1       0.71      0.78      0.74      1000
           2       0.50      0.41      0.45      1000
           3       0.48      0.39      0.43      1000
           4       0.68      0.37      0.48      1000
           5       0.48      0.64      0.54      1000
           6       0.58      0.83      0.69      1000
           7       0.71      0.68      0.69      1000
           8       0.79      0.69      0.74      1000
           9       0.76      0.66      0.71      1000

    accuracy                           0.62     10000
   macro avg       0.63      0.62      0.61     10000
weighted avg       0.63      0.62      0.61     10000

Accuracy Score: 0.6189
Root Mean Square Error: 2.6270896444544865
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-3: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_32'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=32, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 8ms/step - accuracy: 0.3781 - loss: 1.7647 - val_accuracy: 0.5180 - val_loss: 1.3971
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 7ms/step - accuracy: 0.5422 - loss: 1.3151 - val_accuracy: 0.5574 - val_loss: 1.3012
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 8ms/step - accuracy: 0.5786 - loss: 1.2217 - val_accuracy: 0.5876 - val_loss: 1.2057
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 4ms/step - accuracy: 0.6015 - loss: 1.1559 - val_accuracy: 0.5978 - val_loss: 1.1903
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6152 - loss: 1.1115 - val_accuracy: 0.6118 - val_loss: 1.1505
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6343 - loss: 1.0623 - val_accuracy: 0.6238 - val_loss: 1.1262
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 4ms/step - accuracy: 0.6371 - loss: 1.0452 - val_accuracy: 0.6220 - val_loss: 1.1115
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6466 - loss: 1.0200 - val_accuracy: 0.6318 - val_loss: 1.0780
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6568 - loss: 1.0004 - val_accuracy: 0.6344 - val_loss: 1.0656
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 5ms/step - accuracy: 0.6604 - loss: 0.9811 - val_accuracy: 0.6362 - val_loss: 1.0755
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6612 - loss: 0.9781 - val_accuracy: 0.6306 - val_loss: 1.0636
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6705 - loss: 0.9535 - val_accuracy: 0.6400 - val_loss: 1.0451
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6735 - loss: 0.9423 - val_accuracy: 0.6466 - val_loss: 1.0447
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6776 - loss: 0.9265 - val_accuracy: 0.6388 - val_loss: 1.0443
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6758 - loss: 0.9286 - val_accuracy: 0.6564 - val_loss: 1.0149
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6836 - loss: 0.9163 - val_accuracy: 0.6448 - val_loss: 1.0413
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6837 - loss: 0.9107 - val_accuracy: 0.6474 - val_loss: 1.0390
Epoch 18/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6920 - loss: 0.9065 - val_accuracy: 0.6496 - val_loss: 1.0142
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.6436 - loss: 1.0291
313/313 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.62      0.73      0.67      1000
           1       0.76      0.75      0.75      1000
           2       0.48      0.55      0.51      1000
           3       0.45      0.54      0.49      1000
           4       0.55      0.60      0.57      1000
           5       0.65      0.35      0.46      1000
           6       0.74      0.73      0.74      1000
           7       0.78      0.66      0.72      1000
           8       0.74      0.77      0.75      1000
           9       0.74      0.72      0.73      1000

    accuracy                           0.64     10000
   macro avg       0.65      0.64      0.64     10000
weighted avg       0.65      0.64      0.64     10000

Accuracy Score: 0.6401
Root Mean Square Error: 2.5772659932571957
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-4: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_64'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 9s 9ms/step - accuracy: 0.3945 - loss: 1.7022 - val_accuracy: 0.5544 - val_loss: 1.3229
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.5636 - loss: 1.2511 - val_accuracy: 0.5914 - val_loss: 1.1967
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6047 - loss: 1.1435 - val_accuracy: 0.5960 - val_loss: 1.1648
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6292 - loss: 1.0650 - val_accuracy: 0.6190 - val_loss: 1.1017
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6523 - loss: 1.0133 - val_accuracy: 0.6206 - val_loss: 1.1015
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6704 - loss: 0.9642 - val_accuracy: 0.6392 - val_loss: 1.0671
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6786 - loss: 0.9340 - val_accuracy: 0.6450 - val_loss: 1.0479
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6826 - loss: 0.9170 - val_accuracy: 0.6448 - val_loss: 1.0263
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6935 - loss: 0.8880 - val_accuracy: 0.6500 - val_loss: 1.0198
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.7052 - loss: 0.8639 - val_accuracy: 0.6440 - val_loss: 1.0176
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.7041 - loss: 0.8513 - val_accuracy: 0.6448 - val_loss: 1.0176
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.7054 - loss: 0.8507 - val_accuracy: 0.6600 - val_loss: 0.9787
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.7116 - loss: 0.8273 - val_accuracy: 0.6502 - val_loss: 1.0071
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.7166 - loss: 0.8203 - val_accuracy: 0.6522 - val_loss: 1.0018
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.7141 - loss: 0.8151 - val_accuracy: 0.6644 - val_loss: 0.9656
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.7236 - loss: 0.8032 - val_accuracy: 0.6524 - val_loss: 1.0016
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 5ms/step - accuracy: 0.7273 - loss: 0.7774 - val_accuracy: 0.6566 - val_loss: 0.9816
Epoch 18/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.7310 - loss: 0.7799 - val_accuracy: 0.6622 - val_loss: 0.9679
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 5ms/step - accuracy: 0.6562 - loss: 0.9825
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.67      0.65      0.66      1000
           1       0.83      0.70      0.76      1000
           2       0.49      0.54      0.51      1000
           3       0.55      0.40      0.46      1000
           4       0.61      0.61      0.61      1000
           5       0.58      0.53      0.56      1000
           6       0.71      0.80      0.75      1000
           7       0.73      0.73      0.73      1000
           8       0.66      0.83      0.74      1000
           9       0.72      0.77      0.74      1000

    accuracy                           0.66     10000
   macro avg       0.66      0.66      0.65     10000
weighted avg       0.66      0.66      0.65     10000

Accuracy Score: 0.6559
Root Mean Square Error: 2.5302964253225353
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-5: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_128'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 12ms/step - accuracy: 0.4039 - loss: 1.6900 - val_accuracy: 0.5378 - val_loss: 1.3547
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.5700 - loss: 1.2423 - val_accuracy: 0.5920 - val_loss: 1.2093
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6135 - loss: 1.1165 - val_accuracy: 0.6038 - val_loss: 1.1598
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6361 - loss: 1.0426 - val_accuracy: 0.6224 - val_loss: 1.1094
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6596 - loss: 0.9776 - val_accuracy: 0.6268 - val_loss: 1.0803
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6824 - loss: 0.9275 - val_accuracy: 0.6436 - val_loss: 1.0367
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6992 - loss: 0.8805 - val_accuracy: 0.6620 - val_loss: 1.0060
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7084 - loss: 0.8425 - val_accuracy: 0.6368 - val_loss: 1.0383
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7180 - loss: 0.8166 - val_accuracy: 0.6378 - val_loss: 1.0222
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7251 - loss: 0.7920 - val_accuracy: 0.6422 - val_loss: 1.0190
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 0.6511 - loss: 1.0319
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.75      0.60      0.67      1000
           1       0.71      0.83      0.77      1000
           2       0.59      0.37      0.46      1000
           3       0.48      0.46      0.47      1000
           4       0.62      0.58      0.60      1000
           5       0.53      0.61      0.57      1000
           6       0.63      0.84      0.72      1000
           7       0.64      0.79      0.71      1000
           8       0.84      0.68      0.75      1000
           9       0.74      0.74      0.74      1000

    accuracy                           0.65     10000
   macro avg       0.65      0.65      0.64     10000
weighted avg       0.65      0.65      0.64     10000

Accuracy Score: 0.6498
Root Mean Square Error: 2.5028983199482955
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-5: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_256'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 13s 14ms/step - accuracy: 0.4030 - loss: 1.6863 - val_accuracy: 0.5580 - val_loss: 1.2902
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 15s 9ms/step - accuracy: 0.5761 - loss: 1.2200 - val_accuracy: 0.5878 - val_loss: 1.2253
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 9ms/step - accuracy: 0.6100 - loss: 1.1202 - val_accuracy: 0.5844 - val_loss: 1.2006
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 9ms/step - accuracy: 0.6465 - loss: 1.0253 - val_accuracy: 0.6212 - val_loss: 1.1092
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.6759 - loss: 0.9436 - val_accuracy: 0.6446 - val_loss: 1.0457
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.6926 - loss: 0.8823 - val_accuracy: 0.6410 - val_loss: 1.0553
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.7055 - loss: 0.8451 - val_accuracy: 0.6492 - val_loss: 1.0120
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 9ms/step - accuracy: 0.7247 - loss: 0.8000 - val_accuracy: 0.6542 - val_loss: 1.0018
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 9ms/step - accuracy: 0.7295 - loss: 0.7779 - val_accuracy: 0.6352 - val_loss: 1.0549
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 9ms/step - accuracy: 0.7370 - loss: 0.7524 - val_accuracy: 0.6526 - val_loss: 1.0065
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 9ms/step - accuracy: 0.7547 - loss: 0.7054 - val_accuracy: 0.6460 - val_loss: 1.0338
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.6345 - loss: 1.0640
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.71      0.60      0.65      1000
           1       0.80      0.74      0.77      1000
           2       0.43      0.59      0.50      1000
           3       0.47      0.43      0.45      1000
           4       0.61      0.50      0.55      1000
           5       0.56      0.51      0.53      1000
           6       0.61      0.82      0.70      1000
           7       0.79      0.66      0.71      1000
           8       0.82      0.70      0.75      1000
           9       0.67      0.78      0.72      1000

    accuracy                           0.63     10000
   macro avg       0.65      0.63      0.63     10000
weighted avg       0.65      0.63      0.63     10000

Accuracy Score: 0.6334
Root Mean Square Error: 2.5644297611749867
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-7: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_512'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=512, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 19s 22ms/step - accuracy: 0.4036 - loss: 1.7000 - val_accuracy: 0.5482 - val_loss: 1.3003
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 16ms/step - accuracy: 0.5810 - loss: 1.2165 - val_accuracy: 0.5818 - val_loss: 1.2270
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 12s 16ms/step - accuracy: 0.6226 - loss: 1.0928 - val_accuracy: 0.5672 - val_loss: 1.2578
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 20s 16ms/step - accuracy: 0.6486 - loss: 1.0229 - val_accuracy: 0.6124 - val_loss: 1.1316
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 16ms/step - accuracy: 0.6692 - loss: 0.9536 - val_accuracy: 0.6254 - val_loss: 1.0709
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 20s 16ms/step - accuracy: 0.6952 - loss: 0.8885 - val_accuracy: 0.6378 - val_loss: 1.0625
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 16ms/step - accuracy: 0.7101 - loss: 0.8376 - val_accuracy: 0.6408 - val_loss: 1.0425
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 16ms/step - accuracy: 0.7264 - loss: 0.7863 - val_accuracy: 0.6390 - val_loss: 1.0509
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 21s 16ms/step - accuracy: 0.7408 - loss: 0.7500 - val_accuracy: 0.6584 - val_loss: 1.0029
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 20s 16ms/step - accuracy: 0.7531 - loss: 0.7191 - val_accuracy: 0.6370 - val_loss: 1.0672
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 21s 16ms/step - accuracy: 0.7594 - loss: 0.6907 - val_accuracy: 0.6572 - val_loss: 1.0042
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 16ms/step - accuracy: 0.7723 - loss: 0.6594 - val_accuracy: 0.6430 - val_loss: 1.0652
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.6335 - loss: 1.0966
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.66      0.65      0.66      1000
           1       0.69      0.83      0.76      1000
           2       0.51      0.48      0.50      1000
           3       0.46      0.52      0.49      1000
           4       0.65      0.49      0.56      1000
           5       0.62      0.41      0.49      1000
           6       0.65      0.80      0.72      1000
           7       0.73      0.68      0.70      1000
           8       0.78      0.69      0.73      1000
           9       0.63      0.80      0.70      1000

    accuracy                           0.64     10000
   macro avg       0.64      0.64      0.63     10000
weighted avg       0.64      0.64      0.63     10000

Accuracy Score: 0.6353
Root Mean Square Error: 2.63897707454991
No description has been provided for this image
No description has been provided for this image
In [ ]:
data_df = pd.DataFrame(data)
data_df
Out[ ]:
model accuracy val_accuracy test_accuracy loss val_loss test_loss time
0 DNN 0.565 0.562 0.567 1.245 1.274 1.268 46.604
1 CNN_DO_MP_DO_16 0.634 0.610 0.619 1.056 1.110 1.108 60.615
2 CNN_DO_MP_DO_32 0.684 0.650 0.640 0.918 1.014 1.041 85.284
3 CNN_DO_MP_DO_64 0.726 0.662 0.656 0.792 0.968 1.000 82.357
4 CNN_DO_MP_DO_128 0.722 0.642 0.650 0.803 1.019 1.041 55.198
5 CNN_DO_MP_DO_256 0.746 0.646 0.633 0.728 1.034 1.080 106.222
6 CNN_DO_MP_DO_512 0.764 0.643 0.635 0.677 1.065 1.105 200.123
In [ ]:
# Build Model 2-8: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_32-2.2'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=32, kernel_size=(2, 2), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 9s 8ms/step - accuracy: 0.3688 - loss: 1.7767 - val_accuracy: 0.5204 - val_loss: 1.4312
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.5388 - loss: 1.3298 - val_accuracy: 0.5614 - val_loss: 1.3113
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.5710 - loss: 1.2361 - val_accuracy: 0.5774 - val_loss: 1.2541
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.5897 - loss: 1.1860 - val_accuracy: 0.5852 - val_loss: 1.2188
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6034 - loss: 1.1488 - val_accuracy: 0.6058 - val_loss: 1.1890
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6103 - loss: 1.1251 - val_accuracy: 0.6104 - val_loss: 1.1726
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 8s 4ms/step - accuracy: 0.6243 - loss: 1.0918 - val_accuracy: 0.6138 - val_loss: 1.1570
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6315 - loss: 1.0696 - val_accuracy: 0.6188 - val_loss: 1.1385
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.6329 - loss: 1.0644 - val_accuracy: 0.6208 - val_loss: 1.1251
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.6348 - loss: 1.0450 - val_accuracy: 0.6176 - val_loss: 1.1284
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6370 - loss: 1.0509 - val_accuracy: 0.6264 - val_loss: 1.1080
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.6429 - loss: 1.0230 - val_accuracy: 0.6268 - val_loss: 1.1013
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.6532 - loss: 1.0128 - val_accuracy: 0.6220 - val_loss: 1.1030
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 8ms/step - accuracy: 0.6582 - loss: 0.9991 - val_accuracy: 0.6292 - val_loss: 1.0958
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 8s 5ms/step - accuracy: 0.6594 - loss: 0.9873 - val_accuracy: 0.6248 - val_loss: 1.0996
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.6573 - loss: 0.9911 - val_accuracy: 0.6328 - val_loss: 1.0842
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6646 - loss: 0.9759 - val_accuracy: 0.6318 - val_loss: 1.0728
Epoch 18/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6644 - loss: 0.9746 - val_accuracy: 0.6316 - val_loss: 1.0777
Epoch 19/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6615 - loss: 0.9768 - val_accuracy: 0.6316 - val_loss: 1.0770
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.6321 - loss: 1.0671
313/313 ━━━━━━━━━━━━━━━━━━━━ 0s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.72      0.62      0.67      1000
           1       0.74      0.76      0.75      1000
           2       0.51      0.44      0.47      1000
           3       0.48      0.42      0.45      1000
           4       0.58      0.50      0.53      1000
           5       0.62      0.44      0.51      1000
           6       0.58      0.85      0.69      1000
           7       0.72      0.70      0.71      1000
           8       0.64      0.81      0.72      1000
           9       0.67      0.73      0.70      1000

    accuracy                           0.63     10000
   macro avg       0.63      0.63      0.62     10000
weighted avg       0.63      0.63      0.62     10000

Accuracy Score: 0.6277
Root Mean Square Error: 2.6621983397185116
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-9: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_32-4.4'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=32, kernel_size=(4, 4), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 8s 7ms/step - accuracy: 0.3678 - loss: 1.7726 - val_accuracy: 0.5088 - val_loss: 1.4260
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 4ms/step - accuracy: 0.5316 - loss: 1.3425 - val_accuracy: 0.5562 - val_loss: 1.2994
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5678 - loss: 1.2316 - val_accuracy: 0.5690 - val_loss: 1.2554
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5960 - loss: 1.1560 - val_accuracy: 0.5966 - val_loss: 1.1786
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6147 - loss: 1.1066 - val_accuracy: 0.6138 - val_loss: 1.1346
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.6279 - loss: 1.0779 - val_accuracy: 0.6134 - val_loss: 1.1088
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6361 - loss: 1.0476 - val_accuracy: 0.6176 - val_loss: 1.1176
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6427 - loss: 1.0305 - val_accuracy: 0.6220 - val_loss: 1.1074
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6480 - loss: 1.0097 - val_accuracy: 0.6358 - val_loss: 1.0586
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.6611 - loss: 0.9865 - val_accuracy: 0.6260 - val_loss: 1.0675
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.6597 - loss: 0.9787 - val_accuracy: 0.6232 - val_loss: 1.1084
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6670 - loss: 0.9689 - val_accuracy: 0.6370 - val_loss: 1.0510
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 5ms/step - accuracy: 0.6716 - loss: 0.9546 - val_accuracy: 0.6404 - val_loss: 1.0396
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.6669 - loss: 0.9576 - val_accuracy: 0.6356 - val_loss: 1.0357
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6746 - loss: 0.9338 - val_accuracy: 0.6510 - val_loss: 1.0267
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 5ms/step - accuracy: 0.6775 - loss: 0.9375 - val_accuracy: 0.6506 - val_loss: 1.0138
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6800 - loss: 0.9308 - val_accuracy: 0.6436 - val_loss: 1.0460
Epoch 18/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6794 - loss: 0.9294 - val_accuracy: 0.6362 - val_loss: 1.0538
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.6458 - loss: 1.0414
313/313 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.73      0.61      0.67      1000
           1       0.75      0.78      0.76      1000
           2       0.60      0.37      0.46      1000
           3       0.55      0.32      0.40      1000
           4       0.48      0.70      0.57      1000
           5       0.52      0.62      0.57      1000
           6       0.63      0.83      0.72      1000
           7       0.66      0.73      0.69      1000
           8       0.81      0.73      0.77      1000
           9       0.74      0.73      0.73      1000

    accuracy                           0.64     10000
   macro avg       0.65      0.64      0.63     10000
weighted avg       0.65      0.64      0.63     10000

Accuracy Score: 0.6409
Root Mean Square Error: 2.4692306494128897
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-10: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_32-5.5'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=32, kernel_size=(5, 5), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 9s 9ms/step - accuracy: 0.3618 - loss: 1.7763 - val_accuracy: 0.5220 - val_loss: 1.3938
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.5309 - loss: 1.3413 - val_accuracy: 0.5344 - val_loss: 1.3127
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5647 - loss: 1.2488 - val_accuracy: 0.5850 - val_loss: 1.2264
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5970 - loss: 1.1745 - val_accuracy: 0.5922 - val_loss: 1.1701
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6136 - loss: 1.1153 - val_accuracy: 0.5654 - val_loss: 1.2258
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6234 - loss: 1.0813 - val_accuracy: 0.6142 - val_loss: 1.1338
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 5ms/step - accuracy: 0.6355 - loss: 1.0605 - val_accuracy: 0.6282 - val_loss: 1.0941
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.6475 - loss: 1.0169 - val_accuracy: 0.6266 - val_loss: 1.0800
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6534 - loss: 1.0053 - val_accuracy: 0.6324 - val_loss: 1.0710
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6562 - loss: 1.0038 - val_accuracy: 0.6172 - val_loss: 1.0899
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6604 - loss: 0.9815 - val_accuracy: 0.6350 - val_loss: 1.0596
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 5ms/step - accuracy: 0.6694 - loss: 0.9637 - val_accuracy: 0.6418 - val_loss: 1.0313
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6709 - loss: 0.9509 - val_accuracy: 0.6360 - val_loss: 1.0612
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6742 - loss: 0.9415 - val_accuracy: 0.6392 - val_loss: 1.0437
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6789 - loss: 0.9261 - val_accuracy: 0.6386 - val_loss: 1.0426
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.6446 - loss: 1.0463
313/313 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.62      0.70      0.66      1000
           1       0.75      0.79      0.77      1000
           2       0.64      0.36      0.46      1000
           3       0.47      0.46      0.46      1000
           4       0.52      0.66      0.58      1000
           5       0.62      0.46      0.53      1000
           6       0.59      0.85      0.70      1000
           7       0.79      0.68      0.73      1000
           8       0.74      0.75      0.75      1000
           9       0.75      0.70      0.72      1000

    accuracy                           0.64     10000
   macro avg       0.65      0.64      0.64     10000
weighted avg       0.65      0.64      0.64     10000

Accuracy Score: 0.6411
Root Mean Square Error: 2.5542709331627296
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-11: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_64-2.2'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=64, kernel_size=(2, 2), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 13s 8ms/step - accuracy: 0.3773 - loss: 1.7606 - val_accuracy: 0.5342 - val_loss: 1.3675
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 8ms/step - accuracy: 0.5525 - loss: 1.2938 - val_accuracy: 0.5704 - val_loss: 1.2658
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5880 - loss: 1.1903 - val_accuracy: 0.5796 - val_loss: 1.2386
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6140 - loss: 1.1255 - val_accuracy: 0.5898 - val_loss: 1.1889
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 8ms/step - accuracy: 0.6235 - loss: 1.0961 - val_accuracy: 0.6090 - val_loss: 1.1588
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 8ms/step - accuracy: 0.6395 - loss: 1.0452 - val_accuracy: 0.6146 - val_loss: 1.1270
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 9s 7ms/step - accuracy: 0.6409 - loss: 1.0356 - val_accuracy: 0.6122 - val_loss: 1.1421
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6522 - loss: 1.0070 - val_accuracy: 0.5524 - val_loss: 1.2644
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6586 - loss: 0.9967 - val_accuracy: 0.6146 - val_loss: 1.1192
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.6222 - loss: 1.1082
313/313 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.72      0.61      0.66      1000
           1       0.79      0.66      0.72      1000
           2       0.56      0.35      0.43      1000
           3       0.48      0.37      0.41      1000
           4       0.54      0.54      0.54      1000
           5       0.47      0.66      0.55      1000
           6       0.66      0.73      0.69      1000
           7       0.64      0.73      0.68      1000
           8       0.73      0.76      0.74      1000
           9       0.62      0.78      0.69      1000

    accuracy                           0.62     10000
   macro avg       0.62      0.62      0.61     10000
weighted avg       0.62      0.62      0.61     10000

Accuracy Score: 0.6176
Root Mean Square Error: 2.6573483023495434
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-12: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_64-4.4'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=64, kernel_size=(4, 4), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 9ms/step - accuracy: 0.3928 - loss: 1.7050 - val_accuracy: 0.5148 - val_loss: 1.3799
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.5510 - loss: 1.2857 - val_accuracy: 0.5578 - val_loss: 1.2481
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.5981 - loss: 1.1673 - val_accuracy: 0.5666 - val_loss: 1.2722
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6210 - loss: 1.0951 - val_accuracy: 0.6034 - val_loss: 1.1365
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 5ms/step - accuracy: 0.6382 - loss: 1.0484 - val_accuracy: 0.6122 - val_loss: 1.1287
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6522 - loss: 0.9985 - val_accuracy: 0.6214 - val_loss: 1.0773
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6651 - loss: 0.9734 - val_accuracy: 0.6258 - val_loss: 1.0739
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 5ms/step - accuracy: 0.6744 - loss: 0.9484 - val_accuracy: 0.6414 - val_loss: 1.0238
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.6840 - loss: 0.9135 - val_accuracy: 0.6386 - val_loss: 1.0418
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6905 - loss: 0.8962 - val_accuracy: 0.6252 - val_loss: 1.0777
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6953 - loss: 0.8789 - val_accuracy: 0.6208 - val_loss: 1.0809
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.6169 - loss: 1.0932
313/313 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.66      0.65      0.65      1000
           1       0.82      0.69      0.75      1000
           2       0.40      0.64      0.49      1000
           3       0.43      0.49      0.45      1000
           4       0.65      0.39      0.49      1000
           5       0.45      0.66      0.54      1000
           6       0.77      0.69      0.73      1000
           7       0.79      0.60      0.68      1000
           8       0.77      0.73      0.75      1000
           9       0.82      0.61      0.70      1000

    accuracy                           0.61     10000
   macro avg       0.66      0.61      0.62     10000
weighted avg       0.66      0.61      0.62     10000

Accuracy Score: 0.6146
Root Mean Square Error: 2.53327850817868
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-13: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_64-5.5'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=64, kernel_size=(5, 5), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 9s 8ms/step - accuracy: 0.3703 - loss: 1.7598 - val_accuracy: 0.5092 - val_loss: 1.4101
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.5373 - loss: 1.3190 - val_accuracy: 0.5582 - val_loss: 1.2927
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.5796 - loss: 1.2112 - val_accuracy: 0.5734 - val_loss: 1.2255
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6059 - loss: 1.1305 - val_accuracy: 0.5820 - val_loss: 1.1928
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6213 - loss: 1.0856 - val_accuracy: 0.6218 - val_loss: 1.1085
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6428 - loss: 1.0249 - val_accuracy: 0.6302 - val_loss: 1.0709
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6608 - loss: 0.9802 - val_accuracy: 0.6378 - val_loss: 1.0637
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6614 - loss: 0.9704 - val_accuracy: 0.6170 - val_loss: 1.1175
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.6661 - loss: 0.9571 - val_accuracy: 0.6322 - val_loss: 1.0562
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6816 - loss: 0.9235 - val_accuracy: 0.6334 - val_loss: 1.0501
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.6353 - loss: 1.0491
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.66      0.68      0.67      1000
           1       0.81      0.72      0.76      1000
           2       0.52      0.46      0.49      1000
           3       0.47      0.44      0.46      1000
           4       0.45      0.72      0.55      1000
           5       0.65      0.41      0.51      1000
           6       0.77      0.69      0.73      1000
           7       0.68      0.73      0.71      1000
           8       0.81      0.70      0.75      1000
           9       0.66      0.80      0.72      1000

    accuracy                           0.64     10000
   macro avg       0.65      0.64      0.63     10000
weighted avg       0.65      0.64      0.63     10000

Accuracy Score: 0.6352
Root Mean Square Error: 2.5438946519067964
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-14: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_128-2.2'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=128, kernel_size=(2, 2), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 11ms/step - accuracy: 0.3963 - loss: 1.7154 - val_accuracy: 0.5562 - val_loss: 1.3285
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.5651 - loss: 1.2422 - val_accuracy: 0.5564 - val_loss: 1.2822
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6018 - loss: 1.1450 - val_accuracy: 0.5870 - val_loss: 1.2152
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6237 - loss: 1.0816 - val_accuracy: 0.6016 - val_loss: 1.1655
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6460 - loss: 1.0313 - val_accuracy: 0.5770 - val_loss: 1.1958
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6581 - loss: 0.9908 - val_accuracy: 0.5966 - val_loss: 1.1499
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6665 - loss: 0.9667 - val_accuracy: 0.6338 - val_loss: 1.0903
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6736 - loss: 0.9447 - val_accuracy: 0.6330 - val_loss: 1.0787
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6853 - loss: 0.9159 - val_accuracy: 0.6372 - val_loss: 1.0710
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6889 - loss: 0.8906 - val_accuracy: 0.6366 - val_loss: 1.0595
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7028 - loss: 0.8672 - val_accuracy: 0.6364 - val_loss: 1.0660
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7037 - loss: 0.8442 - val_accuracy: 0.6482 - val_loss: 1.0317
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7059 - loss: 0.8413 - val_accuracy: 0.6338 - val_loss: 1.0654
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 7ms/step - accuracy: 0.7213 - loss: 0.8065 - val_accuracy: 0.6466 - val_loss: 1.0276
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7179 - loss: 0.8041 - val_accuracy: 0.6266 - val_loss: 1.0596
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.6287 - loss: 1.0601
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.58      0.76      0.66      1000
           1       0.71      0.82      0.76      1000
           2       0.45      0.54      0.49      1000
           3       0.48      0.40      0.44      1000
           4       0.54      0.58      0.56      1000
           5       0.59      0.46      0.52      1000
           6       0.62      0.83      0.71      1000
           7       0.82      0.61      0.70      1000
           8       0.83      0.64      0.72      1000
           9       0.80      0.63      0.70      1000

    accuracy                           0.63     10000
   macro avg       0.64      0.63      0.63     10000
weighted avg       0.64      0.63      0.63     10000

Accuracy Score: 0.6273
Root Mean Square Error: 2.6697565432076384
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-15: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_128-4.4'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=128, kernel_size=(4,4), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 10ms/step - accuracy: 0.3858 - loss: 1.7222 - val_accuracy: 0.5450 - val_loss: 1.3301
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.5552 - loss: 1.2711 - val_accuracy: 0.5652 - val_loss: 1.2445
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6016 - loss: 1.1490 - val_accuracy: 0.5952 - val_loss: 1.1736
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6257 - loss: 1.0781 - val_accuracy: 0.6084 - val_loss: 1.1465
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6494 - loss: 1.0199 - val_accuracy: 0.6200 - val_loss: 1.1034
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6643 - loss: 0.9652 - val_accuracy: 0.6110 - val_loss: 1.1248
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6765 - loss: 0.9348 - val_accuracy: 0.6354 - val_loss: 1.0642
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6872 - loss: 0.8975 - val_accuracy: 0.6454 - val_loss: 1.0333
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7016 - loss: 0.8639 - val_accuracy: 0.6402 - val_loss: 1.0399
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7130 - loss: 0.8258 - val_accuracy: 0.6472 - val_loss: 1.0304
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7154 - loss: 0.8212 - val_accuracy: 0.6432 - val_loss: 1.0237
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7256 - loss: 0.7910 - val_accuracy: 0.6426 - val_loss: 1.0358
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7315 - loss: 0.7749 - val_accuracy: 0.6586 - val_loss: 1.0101
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7414 - loss: 0.7407 - val_accuracy: 0.6416 - val_loss: 1.0653
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7503 - loss: 0.7314 - val_accuracy: 0.6356 - val_loss: 1.0841
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7524 - loss: 0.7018 - val_accuracy: 0.6374 - val_loss: 1.0786
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 5ms/step - accuracy: 0.6349 - loss: 1.1047
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.57      0.77      0.65      1000
           1       0.71      0.80      0.76      1000
           2       0.54      0.46      0.49      1000
           3       0.47      0.43      0.45      1000
           4       0.64      0.51      0.56      1000
           5       0.57      0.53      0.55      1000
           6       0.62      0.82      0.70      1000
           7       0.81      0.60      0.69      1000
           8       0.74      0.76      0.75      1000
           9       0.73      0.69      0.71      1000

    accuracy                           0.64     10000
   macro avg       0.64      0.64      0.63     10000
weighted avg       0.64      0.64      0.63     10000

Accuracy Score: 0.6359
Root Mean Square Error: 2.633571719167716
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-16: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_128-5.5'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=128, kernel_size=(5,5), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 11ms/step - accuracy: 0.3837 - loss: 1.7397 - val_accuracy: 0.5346 - val_loss: 1.3491
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.5489 - loss: 1.3056 - val_accuracy: 0.5648 - val_loss: 1.2526
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.5898 - loss: 1.1832 - val_accuracy: 0.5834 - val_loss: 1.1942
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6133 - loss: 1.1088 - val_accuracy: 0.5960 - val_loss: 1.1702
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6361 - loss: 1.0453 - val_accuracy: 0.6060 - val_loss: 1.1208
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6578 - loss: 0.9870 - val_accuracy: 0.6150 - val_loss: 1.1127
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.6689 - loss: 0.9546 - val_accuracy: 0.6226 - val_loss: 1.0826
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6803 - loss: 0.9282 - val_accuracy: 0.6360 - val_loss: 1.0738
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6852 - loss: 0.9064 - val_accuracy: 0.6158 - val_loss: 1.0912
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.6972 - loss: 0.8745 - val_accuracy: 0.6094 - val_loss: 1.1263
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7036 - loss: 0.8574 - val_accuracy: 0.6470 - val_loss: 1.0366
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7097 - loss: 0.8321 - val_accuracy: 0.6350 - val_loss: 1.0626
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.7195 - loss: 0.8036 - val_accuracy: 0.6432 - val_loss: 1.0597
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7246 - loss: 0.7944 - val_accuracy: 0.6316 - val_loss: 1.0683
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.6437 - loss: 1.0588
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.62      0.73      0.67      1000
           1       0.80      0.75      0.78      1000
           2       0.58      0.42      0.49      1000
           3       0.50      0.44      0.47      1000
           4       0.51      0.64      0.57      1000
           5       0.58      0.51      0.54      1000
           6       0.59      0.85      0.70      1000
           7       0.72      0.69      0.70      1000
           8       0.84      0.69      0.76      1000
           9       0.76      0.71      0.73      1000

    accuracy                           0.64     10000
   macro avg       0.65      0.64      0.64     10000
weighted avg       0.65      0.64      0.64     10000

Accuracy Score: 0.6424
Root Mean Square Error: 2.5385625854014315
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-17: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_CNN_DO_MP_DO16-2.2'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=16, kernel_size=(3,3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=16, kernel_size=(3,3), strides=(1, 1), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 12s 9ms/step - accuracy: 0.2619 - loss: 2.0019 - val_accuracy: 0.4628 - val_loss: 1.6368
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.4498 - loss: 1.5295 - val_accuracy: 0.5096 - val_loss: 1.4945
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.4899 - loss: 1.4331 - val_accuracy: 0.5270 - val_loss: 1.4198
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.5216 - loss: 1.3540 - val_accuracy: 0.5526 - val_loss: 1.3674
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.5293 - loss: 1.3230 - val_accuracy: 0.5666 - val_loss: 1.3325
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.5464 - loss: 1.2886 - val_accuracy: 0.5856 - val_loss: 1.2964
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.5544 - loss: 1.2610 - val_accuracy: 0.5894 - val_loss: 1.2844
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.5592 - loss: 1.2432 - val_accuracy: 0.5710 - val_loss: 1.2728
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.5660 - loss: 1.2300 - val_accuracy: 0.6018 - val_loss: 1.2442
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 5ms/step - accuracy: 0.5669 - loss: 1.2162 - val_accuracy: 0.6044 - val_loss: 1.2158
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5721 - loss: 1.2129 - val_accuracy: 0.5858 - val_loss: 1.2454
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5763 - loss: 1.1965 - val_accuracy: 0.6088 - val_loss: 1.2181
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5795 - loss: 1.1938 - val_accuracy: 0.6218 - val_loss: 1.1841
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5847 - loss: 1.1775 - val_accuracy: 0.5988 - val_loss: 1.1922
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5880 - loss: 1.1794 - val_accuracy: 0.6134 - val_loss: 1.1736
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5908 - loss: 1.1609 - val_accuracy: 0.6168 - val_loss: 1.1839
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 5ms/step - accuracy: 0.6279 - loss: 1.1887
313/313 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.67      0.65      0.66      1000
           1       0.78      0.74      0.76      1000
           2       0.55      0.45      0.49      1000
           3       0.46      0.48      0.47      1000
           4       0.52      0.63      0.57      1000
           5       0.56      0.51      0.54      1000
           6       0.64      0.81      0.72      1000
           7       0.78      0.61      0.69      1000
           8       0.62      0.81      0.70      1000
           9       0.83      0.60      0.70      1000

    accuracy                           0.63     10000
   macro avg       0.64      0.63      0.63     10000
weighted avg       0.64      0.63      0.63     10000

Accuracy Score: 0.6298
Root Mean Square Error: 2.5800775182152957
No description has been provided for this image
No description has been provided for this image
In [ ]:
data_df = pd.DataFrame(data)
data_df
Out[ ]:
model accuracy val_accuracy test_accuracy loss val_loss test_loss time
0 DNN 0.565 0.562 0.567 1.245 1.274 1.268 46.604
1 CNN_DO_MP_DO_16 0.634 0.610 0.619 1.056 1.110 1.108 60.615
2 CNN_DO_MP_DO_32 0.684 0.650 0.640 0.918 1.014 1.041 85.284
3 CNN_DO_MP_DO_64 0.726 0.662 0.656 0.792 0.968 1.000 82.357
4 CNN_DO_MP_DO_128 0.722 0.642 0.650 0.803 1.019 1.041 55.198
5 CNN_DO_MP_DO_256 0.746 0.646 0.633 0.728 1.034 1.080 106.222
6 CNN_DO_MP_DO_512 0.764 0.643 0.635 0.677 1.065 1.105 200.123
7 CNN_DO_MP_DO_32-2.2 0.662 0.632 0.628 0.978 1.077 1.080 100.190
8 CNN_DO_MP_DO_32-4.4 0.675 0.636 0.641 0.938 1.054 1.053 74.688
9 CNN_DO_MP_DO_32-5.5 0.674 0.639 0.641 0.943 1.043 1.058 68.455
10 CNN_DO_MP_DO_64-2.2 0.655 0.615 0.618 1.003 1.119 1.122 63.568
11 CNN_DO_MP_DO_64-4.4 0.691 0.621 0.615 0.894 1.081 1.113 58.112
12 CNN_DO_MP_DO_64-5.5 0.676 0.633 0.635 0.938 1.050 1.062 45.898
13 CNN_DO_MP_DO_128-2.2 0.715 0.627 0.627 0.814 1.060 1.074 79.575
14 CNN_DO_MP_DO_128-4.4 0.746 0.637 0.636 0.724 1.079 1.110 82.708
15 CNN_DO_MP_DO_128-5.5 0.721 0.632 0.642 0.807 1.068 1.075 73.661
16 CNN_DO_MP_DO_CNN_DO_MP_DO16-2.2 0.590 0.617 0.630 1.167 1.184 1.189 75.247
In [ ]:
# Build Model 2-4: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_CNN_DO_MP_DO_64_64'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 16s 17ms/step - accuracy: 0.3441 - loss: 1.7953 - val_accuracy: 0.5262 - val_loss: 1.4136
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 6ms/step - accuracy: 0.5422 - loss: 1.2989 - val_accuracy: 0.5820 - val_loss: 1.2772
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.5962 - loss: 1.1523 - val_accuracy: 0.6346 - val_loss: 1.1458
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6250 - loss: 1.0767 - val_accuracy: 0.6340 - val_loss: 1.1225
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.6400 - loss: 1.0259 - val_accuracy: 0.6640 - val_loss: 1.0534
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.6578 - loss: 0.9897 - val_accuracy: 0.6716 - val_loss: 1.0109
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6695 - loss: 0.9533 - val_accuracy: 0.6766 - val_loss: 0.9946
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.6747 - loss: 0.9251 - val_accuracy: 0.6616 - val_loss: 1.0050
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.6856 - loss: 0.9081 - val_accuracy: 0.6850 - val_loss: 0.9449
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.6900 - loss: 0.8935 - val_accuracy: 0.6916 - val_loss: 0.9607
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.6978 - loss: 0.8687 - val_accuracy: 0.6966 - val_loss: 0.9219
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.6968 - loss: 0.8719 - val_accuracy: 0.7054 - val_loss: 0.9161
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.6973 - loss: 0.8670 - val_accuracy: 0.7156 - val_loss: 0.8970
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.7065 - loss: 0.8388 - val_accuracy: 0.7154 - val_loss: 0.9180
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 5ms/step - accuracy: 0.7078 - loss: 0.8429 - val_accuracy: 0.7092 - val_loss: 0.9020
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.7122 - loss: 0.8271 - val_accuracy: 0.7114 - val_loss: 0.9123
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.7069 - loss: 0.9190
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.76      0.71      0.74      1000
           1       0.89      0.77      0.83      1000
           2       0.62      0.58      0.60      1000
           3       0.57      0.42      0.49      1000
           4       0.60      0.70      0.65      1000
           5       0.61      0.67      0.64      1000
           6       0.63      0.89      0.74      1000
           7       0.86      0.66      0.75      1000
           8       0.80      0.83      0.81      1000
           9       0.78      0.80      0.79      1000

    accuracy                           0.70     10000
   macro avg       0.71      0.70      0.70     10000
weighted avg       0.71      0.70      0.70     10000

Accuracy Score: 0.7042
Root Mean Square Error: 2.2513551474611906
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-4: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_CNN_DO_MP_DO_64_128'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 13s 12ms/step - accuracy: 0.3676 - loss: 1.7470 - val_accuracy: 0.5384 - val_loss: 1.4043
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.5711 - loss: 1.2219 - val_accuracy: 0.5818 - val_loss: 1.2311
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.6243 - loss: 1.0822 - val_accuracy: 0.6524 - val_loss: 1.0748
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6572 - loss: 0.9890 - val_accuracy: 0.6560 - val_loss: 1.0504
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6798 - loss: 0.9328 - val_accuracy: 0.6824 - val_loss: 0.9780
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6947 - loss: 0.8840 - val_accuracy: 0.6750 - val_loss: 0.9667
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7024 - loss: 0.8646 - val_accuracy: 0.6772 - val_loss: 0.9514
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.7136 - loss: 0.8309 - val_accuracy: 0.7168 - val_loss: 0.8910
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 6ms/step - accuracy: 0.7195 - loss: 0.8069 - val_accuracy: 0.7042 - val_loss: 0.8997
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7272 - loss: 0.7896 - val_accuracy: 0.7186 - val_loss: 0.8802
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.7345 - loss: 0.7686 - val_accuracy: 0.7302 - val_loss: 0.8580
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 6ms/step - accuracy: 0.7353 - loss: 0.7591 - val_accuracy: 0.7226 - val_loss: 0.8681
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.7422 - loss: 0.7402 - val_accuracy: 0.7186 - val_loss: 0.8349
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.7448 - loss: 0.7323 - val_accuracy: 0.7088 - val_loss: 0.8800
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.7122 - loss: 0.8789
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.83      0.64      0.72      1000
           1       0.91      0.78      0.84      1000
           2       0.66      0.54      0.59      1000
           3       0.52      0.55      0.53      1000
           4       0.54      0.80      0.65      1000
           5       0.71      0.54      0.61      1000
           6       0.70      0.86      0.77      1000
           7       0.87      0.66      0.75      1000
           8       0.75      0.89      0.81      1000
           9       0.77      0.84      0.80      1000

    accuracy                           0.71     10000
   macro avg       0.73      0.71      0.71     10000
weighted avg       0.73      0.71      0.71     10000

Accuracy Score: 0.7084
Root Mean Square Error: 2.224792125120907
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-4: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_CNN_DO_MP_DO_64_256'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 14s 12ms/step - accuracy: 0.3696 - loss: 1.7344 - val_accuracy: 0.5622 - val_loss: 1.3121
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5910 - loss: 1.1695 - val_accuracy: 0.6002 - val_loss: 1.1910
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6473 - loss: 1.0187 - val_accuracy: 0.6588 - val_loss: 1.0320
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6713 - loss: 0.9528 - val_accuracy: 0.6878 - val_loss: 0.9660
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7001 - loss: 0.8726 - val_accuracy: 0.6802 - val_loss: 0.9566
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7163 - loss: 0.8179 - val_accuracy: 0.7128 - val_loss: 0.9227
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7281 - loss: 0.7816 - val_accuracy: 0.7000 - val_loss: 0.9095
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7400 - loss: 0.7505 - val_accuracy: 0.6974 - val_loss: 0.9004
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7488 - loss: 0.7284 - val_accuracy: 0.7298 - val_loss: 0.8405
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7564 - loss: 0.7050 - val_accuracy: 0.7314 - val_loss: 0.7998
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7663 - loss: 0.6756 - val_accuracy: 0.7372 - val_loss: 0.8064
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7708 - loss: 0.6624 - val_accuracy: 0.7200 - val_loss: 0.8165
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 7ms/step - accuracy: 0.7743 - loss: 0.6441 - val_accuracy: 0.6894 - val_loss: 0.8961
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7781 - loss: 0.6347 - val_accuracy: 0.7402 - val_loss: 0.7693
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7820 - loss: 0.6149 - val_accuracy: 0.7298 - val_loss: 0.8037
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7918 - loss: 0.5996 - val_accuracy: 0.7396 - val_loss: 0.7716
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7921 - loss: 0.5946 - val_accuracy: 0.7398 - val_loss: 0.7775
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.7462 - loss: 0.7807
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.83      0.72      0.77      1000
           1       0.88      0.84      0.86      1000
           2       0.71      0.59      0.64      1000
           3       0.63      0.49      0.55      1000
           4       0.64      0.76      0.69      1000
           5       0.64      0.69      0.66      1000
           6       0.70      0.87      0.78      1000
           7       0.77      0.78      0.78      1000
           8       0.83      0.86      0.84      1000
           9       0.81      0.83      0.82      1000

    accuracy                           0.74     10000
   macro avg       0.74      0.74      0.74     10000
weighted avg       0.74      0.74      0.74     10000

Accuracy Score: 0.7416
Root Mean Square Error: 2.0865761428713787
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-4: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_X3_128_256_512'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=512, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 27s 26ms/step - accuracy: 0.3281 - loss: 1.8146 - val_accuracy: 0.5386 - val_loss: 1.4395
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 15ms/step - accuracy: 0.5617 - loss: 1.2437 - val_accuracy: 0.6482 - val_loss: 1.1671
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 14ms/step - accuracy: 0.6206 - loss: 1.0864 - val_accuracy: 0.6746 - val_loss: 1.0786
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 14ms/step - accuracy: 0.6689 - loss: 0.9509 - val_accuracy: 0.6918 - val_loss: 1.0420
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 15ms/step - accuracy: 0.6868 - loss: 0.8937 - val_accuracy: 0.7172 - val_loss: 0.9329
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 21s 16ms/step - accuracy: 0.7055 - loss: 0.8408 - val_accuracy: 0.7106 - val_loss: 0.9259
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 19s 15ms/step - accuracy: 0.7175 - loss: 0.8006 - val_accuracy: 0.7254 - val_loss: 0.8725
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 20s 14ms/step - accuracy: 0.7325 - loss: 0.7658 - val_accuracy: 0.7394 - val_loss: 0.8195
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 14ms/step - accuracy: 0.7430 - loss: 0.7347 - val_accuracy: 0.7266 - val_loss: 0.8372
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 14ms/step - accuracy: 0.7506 - loss: 0.7092 - val_accuracy: 0.6950 - val_loss: 0.9184
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 14ms/step - accuracy: 0.7537 - loss: 0.7029 - val_accuracy: 0.7212 - val_loss: 0.8335
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 0.7213 - loss: 0.8511
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.79      0.79      0.79      1000
           1       0.89      0.84      0.86      1000
           2       0.58      0.66      0.62      1000
           3       0.63      0.43      0.51      1000
           4       0.64      0.73      0.68      1000
           5       0.77      0.46      0.58      1000
           6       0.57      0.91      0.70      1000
           7       0.84      0.71      0.77      1000
           8       0.74      0.92      0.82      1000
           9       0.91      0.73      0.81      1000

    accuracy                           0.72     10000
   macro avg       0.74      0.72      0.71     10000
weighted avg       0.74      0.72      0.71     10000

Accuracy Score: 0.7176
Root Mean Square Error: 2.1007855673533173
No description has been provided for this image
No description has been provided for this image
In [ ]:
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
Out[ ]:
No description has been provided for this image
In [ ]:
data_df = pd.DataFrame(data)
data_df
Out[ ]:
model accuracy val_accuracy test_accuracy loss val_loss test_loss time
0 DNN 0.565 0.562 0.567 1.245 1.274 1.268 46.604
1 CNN_DO_MP_DO_16 0.634 0.610 0.619 1.056 1.110 1.108 60.615
2 CNN_DO_MP_DO_32 0.684 0.650 0.640 0.918 1.014 1.041 85.284
3 CNN_DO_MP_DO_64 0.726 0.662 0.656 0.792 0.968 1.000 82.357
4 CNN_DO_MP_DO_128 0.722 0.642 0.650 0.803 1.019 1.041 55.198
5 CNN_DO_MP_DO_256 0.746 0.646 0.633 0.728 1.034 1.080 106.222
6 CNN_DO_MP_DO_512 0.764 0.643 0.635 0.677 1.065 1.105 200.123
7 CNN_DO_MP_DO_32-2.2 0.662 0.632 0.628 0.978 1.077 1.080 100.190
8 CNN_DO_MP_DO_32-4.4 0.675 0.636 0.641 0.938 1.054 1.053 74.688
9 CNN_DO_MP_DO_32-5.5 0.674 0.639 0.641 0.943 1.043 1.058 68.455
10 CNN_DO_MP_DO_64-2.2 0.655 0.615 0.618 1.003 1.119 1.122 63.568
11 CNN_DO_MP_DO_64-4.4 0.691 0.621 0.615 0.894 1.081 1.113 58.112
12 CNN_DO_MP_DO_64-5.5 0.676 0.633 0.635 0.938 1.050 1.062 45.898
13 CNN_DO_MP_DO_128-2.2 0.715 0.627 0.627 0.814 1.060 1.074 79.575
14 CNN_DO_MP_DO_128-4.4 0.746 0.637 0.636 0.724 1.079 1.110 82.708
15 CNN_DO_MP_DO_128-5.5 0.721 0.632 0.642 0.807 1.068 1.075 73.661
16 CNN_DO_MP_DO_CNN_DO_MP_DO16-2.2 0.590 0.617 0.630 1.167 1.184 1.189 75.247
17 CNN_DO_MP_DO_CNN_DO_MP_DO_64_64 0.703 0.688 0.690 0.858 0.939 0.940 70.261
18 CNN_DO_MP_DO_CNN_DO_MP_DO_64_128 0.752 0.723 0.733 0.714 0.815 0.822 107.098
19 CNN_DO_MP_DO_CNN_DO_MP_DO_64_256 0.781 0.741 0.744 0.629 0.789 0.793 111.724
20 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.732 0.726 0.730 0.772 0.813 0.822 71.486
21 CNN_DO_MP_DO_CNN_DO_MP_DO_64_64 0.709 0.711 0.704 0.838 0.912 0.923 88.232
22 CNN_DO_MP_DO_CNN_DO_MP_DO_64_128 0.745 0.709 0.708 0.735 0.880 0.887 76.367
23 CNN_DO_MP_DO_CNN_DO_MP_DO_64_256 0.788 0.740 0.742 0.609 0.777 0.787 100.630
24 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.737 0.713 0.711 0.758 0.856 0.877 75.341
25 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.778 0.732 0.740 0.639 0.763 0.769 136.856
26 CNN_DO_MP_DO_X3_128_256_512 0.753 0.721 0.718 0.702 0.833 0.853 161.721
In [ ]:
# Build Model 2-4: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_X3_64_128_256'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1),activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 23s 19ms/step - accuracy: 0.3181 - loss: 1.8331 - val_accuracy: 0.5126 - val_loss: 1.5171
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5373 - loss: 1.3018 - val_accuracy: 0.6064 - val_loss: 1.2708
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5944 - loss: 1.1549 - val_accuracy: 0.6412 - val_loss: 1.1572
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6307 - loss: 1.0467 - val_accuracy: 0.6742 - val_loss: 1.0769
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6560 - loss: 0.9872 - val_accuracy: 0.6806 - val_loss: 1.0088
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6736 - loss: 0.9372 - val_accuracy: 0.7130 - val_loss: 1.0116
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6900 - loss: 0.8882 - val_accuracy: 0.7146 - val_loss: 0.9262
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7019 - loss: 0.8520 - val_accuracy: 0.7016 - val_loss: 0.9460
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7103 - loss: 0.8327 - val_accuracy: 0.7024 - val_loss: 0.9153
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7191 - loss: 0.8067 - val_accuracy: 0.7338 - val_loss: 0.8750
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7269 - loss: 0.7934 - val_accuracy: 0.7320 - val_loss: 0.8843
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7292 - loss: 0.7686 - val_accuracy: 0.7476 - val_loss: 0.8055
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7376 - loss: 0.7503 - val_accuracy: 0.7452 - val_loss: 0.8051
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7367 - loss: 0.7452 - val_accuracy: 0.7534 - val_loss: 0.7829
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7434 - loss: 0.7379 - val_accuracy: 0.7522 - val_loss: 0.7883
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7501 - loss: 0.7142 - val_accuracy: 0.7392 - val_loss: 0.7972
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7535 - loss: 0.7012 - val_accuracy: 0.7614 - val_loss: 0.7688
Epoch 18/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7576 - loss: 0.6982 - val_accuracy: 0.7652 - val_loss: 0.7429
Epoch 19/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7567 - loss: 0.6983 - val_accuracy: 0.7656 - val_loss: 0.7353
Epoch 20/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7638 - loss: 0.6777 - val_accuracy: 0.7506 - val_loss: 0.7515
Epoch 21/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 8ms/step - accuracy: 0.7652 - loss: 0.6736 - val_accuracy: 0.7594 - val_loss: 0.7291
Epoch 22/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 9s 6ms/step - accuracy: 0.7686 - loss: 0.6561 - val_accuracy: 0.7610 - val_loss: 0.7190
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 5ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.7504 - loss: 0.7494
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.85      0.72      0.78      1000
           1       0.92      0.86      0.89      1000
           2       0.68      0.60      0.64      1000
           3       0.63      0.57      0.60      1000
           4       0.60      0.83      0.70      1000
           5       0.78      0.58      0.67      1000
           6       0.63      0.94      0.76      1000
           7       0.89      0.71      0.79      1000
           8       0.80      0.89      0.84      1000
           9       0.88      0.82      0.85      1000

    accuracy                           0.75     10000
   macro avg       0.77      0.75      0.75     10000
weighted avg       0.77      0.75      0.75     10000

Accuracy Score: 0.7515
Root Mean Square Error: 1.961479033790573
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-4: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_X3_32_64_128'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=32, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1),activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 16s 12ms/step - accuracy: 0.2883 - loss: 1.9172 - val_accuracy: 0.5010 - val_loss: 1.5772
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 9s 5ms/step - accuracy: 0.5022 - loss: 1.3854 - val_accuracy: 0.5662 - val_loss: 1.3735
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5559 - loss: 1.2513 - val_accuracy: 0.6006 - val_loss: 1.2827
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.5831 - loss: 1.1709 - val_accuracy: 0.6274 - val_loss: 1.1893
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.6091 - loss: 1.1116 - val_accuracy: 0.6524 - val_loss: 1.1418
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.6272 - loss: 1.0584 - val_accuracy: 0.6334 - val_loss: 1.1236
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.6374 - loss: 1.0392 - val_accuracy: 0.6304 - val_loss: 1.1232
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 4ms/step - accuracy: 0.6517 - loss: 0.9970 - val_accuracy: 0.6884 - val_loss: 1.0392
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.6526 - loss: 0.9814 - val_accuracy: 0.6578 - val_loss: 1.0467
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.6606 - loss: 0.9668 - val_accuracy: 0.6822 - val_loss: 0.9942
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6722 - loss: 0.9359 - val_accuracy: 0.7016 - val_loss: 1.0022
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6783 - loss: 0.9274 - val_accuracy: 0.6920 - val_loss: 0.9720
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.6820 - loss: 0.9149 - val_accuracy: 0.6836 - val_loss: 0.9839
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.6894 - loss: 0.8923 - val_accuracy: 0.7222 - val_loss: 0.9320
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6839 - loss: 0.8978 - val_accuracy: 0.6848 - val_loss: 0.9624
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 5ms/step - accuracy: 0.6952 - loss: 0.8758 - val_accuracy: 0.6836 - val_loss: 0.9239
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.6916 - loss: 0.8813 - val_accuracy: 0.7282 - val_loss: 0.8712
Epoch 18/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 5ms/step - accuracy: 0.6925 - loss: 0.8679 - val_accuracy: 0.7198 - val_loss: 0.8608
Epoch 19/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.7027 - loss: 0.8450 - val_accuracy: 0.6966 - val_loss: 0.9245
Epoch 20/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.7043 - loss: 0.8439 - val_accuracy: 0.7130 - val_loss: 0.8731
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.7179 - loss: 0.8845
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.88      0.65      0.75      1000
           1       0.92      0.77      0.84      1000
           2       0.66      0.56      0.61      1000
           3       0.60      0.49      0.54      1000
           4       0.57      0.81      0.67      1000
           5       0.70      0.60      0.65      1000
           6       0.58      0.91      0.71      1000
           7       0.82      0.71      0.76      1000
           8       0.80      0.87      0.83      1000
           9       0.83      0.82      0.83      1000

    accuracy                           0.72     10000
   macro avg       0.74      0.72      0.72     10000
weighted avg       0.74      0.72      0.72     10000

Accuracy Score: 0.7181
Root Mean Square Error: 2.1209667607013554
No description has been provided for this image
No description has been provided for this image
In [ ]:
data_df = pd.DataFrame(data)
data_df
Out[ ]:
model accuracy val_accuracy test_accuracy loss val_loss test_loss time
0 DNN 0.565 0.562 0.567 1.245 1.274 1.268 46.604
1 CNN_DO_MP_DO_16 0.634 0.610 0.619 1.056 1.110 1.108 60.615
2 CNN_DO_MP_DO_32 0.684 0.650 0.640 0.918 1.014 1.041 85.284
3 CNN_DO_MP_DO_64 0.726 0.662 0.656 0.792 0.968 1.000 82.357
4 CNN_DO_MP_DO_128 0.722 0.642 0.650 0.803 1.019 1.041 55.198
5 CNN_DO_MP_DO_256 0.746 0.646 0.633 0.728 1.034 1.080 106.222
6 CNN_DO_MP_DO_512 0.764 0.643 0.635 0.677 1.065 1.105 200.123
7 CNN_DO_MP_DO_32-2.2 0.662 0.632 0.628 0.978 1.077 1.080 100.190
8 CNN_DO_MP_DO_32-4.4 0.675 0.636 0.641 0.938 1.054 1.053 74.688
9 CNN_DO_MP_DO_32-5.5 0.674 0.639 0.641 0.943 1.043 1.058 68.455
10 CNN_DO_MP_DO_64-2.2 0.655 0.615 0.618 1.003 1.119 1.122 63.568
11 CNN_DO_MP_DO_64-4.4 0.691 0.621 0.615 0.894 1.081 1.113 58.112
12 CNN_DO_MP_DO_64-5.5 0.676 0.633 0.635 0.938 1.050 1.062 45.898
13 CNN_DO_MP_DO_128-2.2 0.715 0.627 0.627 0.814 1.060 1.074 79.575
14 CNN_DO_MP_DO_128-4.4 0.746 0.637 0.636 0.724 1.079 1.110 82.708
15 CNN_DO_MP_DO_128-5.5 0.721 0.632 0.642 0.807 1.068 1.075 73.661
16 CNN_DO_MP_DO_CNN_DO_MP_DO16-2.2 0.590 0.617 0.630 1.167 1.184 1.189 75.247
17 CNN_DO_MP_DO_CNN_DO_MP_DO_64_64 0.703 0.688 0.690 0.858 0.939 0.940 70.261
18 CNN_DO_MP_DO_CNN_DO_MP_DO_64_128 0.752 0.723 0.733 0.714 0.815 0.822 107.098
19 CNN_DO_MP_DO_CNN_DO_MP_DO_64_256 0.781 0.741 0.744 0.629 0.789 0.793 111.724
20 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.732 0.726 0.730 0.772 0.813 0.822 71.486
21 CNN_DO_MP_DO_CNN_DO_MP_DO_64_64 0.709 0.711 0.704 0.838 0.912 0.923 88.232
22 CNN_DO_MP_DO_CNN_DO_MP_DO_64_128 0.745 0.709 0.708 0.735 0.880 0.887 76.367
23 CNN_DO_MP_DO_CNN_DO_MP_DO_64_256 0.788 0.740 0.742 0.609 0.777 0.787 100.630
24 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.737 0.713 0.711 0.758 0.856 0.877 75.341
25 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.778 0.732 0.740 0.639 0.763 0.769 136.856
26 CNN_DO_MP_DO_X3_128_256_512 0.753 0.721 0.718 0.702 0.833 0.853 161.721
27 CNN_DO_MP_DO_X3_64_128_256 0.765 0.761 0.752 0.666 0.719 0.748 133.832
28 CNN_DO_MP_DO_X3_32_64_128 0.703 0.713 0.718 0.847 0.873 0.884 102.764
In [ ]:
# Build Model 2-4: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_X3_64_128_256-1'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1),activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 16s 14ms/step - accuracy: 0.3341 - loss: 1.7942 - val_accuracy: 0.5576 - val_loss: 1.4894
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5479 - loss: 1.2744 - val_accuracy: 0.6188 - val_loss: 1.2519
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6034 - loss: 1.1208 - val_accuracy: 0.6562 - val_loss: 1.1637
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6377 - loss: 1.0297 - val_accuracy: 0.6894 - val_loss: 1.0645
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6699 - loss: 0.9572 - val_accuracy: 0.6980 - val_loss: 1.0102
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6791 - loss: 0.9193 - val_accuracy: 0.6716 - val_loss: 1.0189
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6915 - loss: 0.8792 - val_accuracy: 0.7208 - val_loss: 0.8952
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7026 - loss: 0.8463 - val_accuracy: 0.7174 - val_loss: 0.8994
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7189 - loss: 0.8148 - val_accuracy: 0.7450 - val_loss: 0.8631
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 8ms/step - accuracy: 0.7257 - loss: 0.7869 - val_accuracy: 0.7198 - val_loss: 0.8711
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7326 - loss: 0.7659 - val_accuracy: 0.7226 - val_loss: 0.8606
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7346 - loss: 0.7603 - val_accuracy: 0.7182 - val_loss: 0.8374
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.7191 - loss: 0.8612
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.86      0.69      0.77      1000
           1       0.95      0.76      0.85      1000
           2       0.61      0.62      0.62      1000
           3       0.50      0.59      0.54      1000
           4       0.67      0.72      0.70      1000
           5       0.68      0.56      0.61      1000
           6       0.58      0.92      0.71      1000
           7       0.92      0.64      0.76      1000
           8       0.82      0.86      0.84      1000
           9       0.85      0.79      0.82      1000

    accuracy                           0.72     10000
   macro avg       0.74      0.72      0.72     10000
weighted avg       0.74      0.72      0.72     10000

Accuracy Score: 0.7162
Root Mean Square Error: 2.0617953341687434
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-4: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_X3_64_128_256-2'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1),activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 17s 14ms/step - accuracy: 0.3142 - loss: 1.8428 - val_accuracy: 0.5426 - val_loss: 1.4761
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5375 - loss: 1.3014 - val_accuracy: 0.5802 - val_loss: 1.3091
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.5954 - loss: 1.1498 - val_accuracy: 0.6390 - val_loss: 1.1598
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6356 - loss: 1.0460 - val_accuracy: 0.6600 - val_loss: 1.1073
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6556 - loss: 0.9871 - val_accuracy: 0.6774 - val_loss: 1.0242
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6724 - loss: 0.9385 - val_accuracy: 0.6832 - val_loss: 1.0019
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6856 - loss: 0.8980 - val_accuracy: 0.7110 - val_loss: 0.9578
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6967 - loss: 0.8664 - val_accuracy: 0.7044 - val_loss: 0.9627
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7087 - loss: 0.8373 - val_accuracy: 0.7280 - val_loss: 0.9038
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7171 - loss: 0.8117 - val_accuracy: 0.7386 - val_loss: 0.8762
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7209 - loss: 0.7994 - val_accuracy: 0.7136 - val_loss: 0.9047
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 7ms/step - accuracy: 0.7224 - loss: 0.7972 - val_accuracy: 0.7334 - val_loss: 0.8688
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7339 - loss: 0.7641 - val_accuracy: 0.7410 - val_loss: 0.8072
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7356 - loss: 0.7513 - val_accuracy: 0.7400 - val_loss: 0.8029
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7413 - loss: 0.7500 - val_accuracy: 0.7458 - val_loss: 0.8300
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7494 - loss: 0.7211 - val_accuracy: 0.7404 - val_loss: 0.7801
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 7ms/step - accuracy: 0.7510 - loss: 0.7175 - val_accuracy: 0.7012 - val_loss: 0.8612
Epoch 18/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7519 - loss: 0.7092 - val_accuracy: 0.7446 - val_loss: 0.7759
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.7418 - loss: 0.7846
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.83      0.73      0.78      1000
           1       0.93      0.81      0.87      1000
           2       0.69      0.58      0.63      1000
           3       0.64      0.46      0.54      1000
           4       0.55      0.86      0.67      1000
           5       0.74      0.60      0.66      1000
           6       0.70      0.89      0.78      1000
           7       0.86      0.74      0.80      1000
           8       0.73      0.93      0.82      1000
           9       0.89      0.78      0.83      1000

    accuracy                           0.74     10000
   macro avg       0.75      0.74      0.74     10000
weighted avg       0.75      0.74      0.74     10000

Accuracy Score: 0.739
Root Mean Square Error: 2.018316129846858
No description has been provided for this image
No description has been provided for this image
In [ ]:
# Build Model 2-4: CNN and MaxPool with dropouts between and after
name = 'CNN_DO_MP_DO_X3_64_128_256-DNN'
k.clear_session()
model = models.Sequential()
model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1),activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.MaxPool2D((2, 2),strides=2))
model.add(layers.Dropout(0.25))
model.add(layers.Dense(units=128, activation=tf.nn.relu))
model.add(layers.Dropout(0.25))
model.add(layers.Flatten())
model.add(layers.Dense(units=10, activation=tf.nn.softmax))
keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
time_start = time.time()
history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                     ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
time_end = time.time()
preds = model.predict(image_test_norm)
test_pred = model.evaluate(image_test_norm, test_labels)

history_dict = history.history
history_df=pd.DataFrame(history_dict)
plt.subplots(figsize=(16,12))
plt.tight_layout()
display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
pred= model.predict(image_test_norm)
pred=np.argmax(pred, axis=1)
print_validation_report(test_labels, pred)
plot_confusion_matrix(test_labels, pred)
add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 27s 18ms/step - accuracy: 0.2668 - loss: 1.9471 - val_accuracy: 0.4806 - val_loss: 1.5537
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 20s 7ms/step - accuracy: 0.5027 - loss: 1.3920 - val_accuracy: 0.5710 - val_loss: 1.3219
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5660 - loss: 1.2126 - val_accuracy: 0.6152 - val_loss: 1.1970
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5979 - loss: 1.1351 - val_accuracy: 0.5774 - val_loss: 1.2030
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6311 - loss: 1.0413 - val_accuracy: 0.5824 - val_loss: 1.1998
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6499 - loss: 0.9863 - val_accuracy: 0.6576 - val_loss: 0.9886
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6707 - loss: 0.9379 - val_accuracy: 0.6982 - val_loss: 0.9347
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6821 - loss: 0.9041 - val_accuracy: 0.7122 - val_loss: 0.8631
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 7ms/step - accuracy: 0.6890 - loss: 0.8909 - val_accuracy: 0.6118 - val_loss: 1.0908
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 7ms/step - accuracy: 0.7022 - loss: 0.8497 - val_accuracy: 0.7148 - val_loss: 0.8302
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7069 - loss: 0.8364 - val_accuracy: 0.7014 - val_loss: 0.8893
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7076 - loss: 0.8298 - val_accuracy: 0.7058 - val_loss: 0.8393
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7174 - loss: 0.8061 - val_accuracy: 0.7034 - val_loss: 0.8497
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 0.7086 - loss: 0.8630
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.88      0.58      0.70      1000
           1       0.91      0.84      0.88      1000
           2       0.66      0.53      0.59      1000
           3       0.61      0.43      0.51      1000
           4       0.52      0.81      0.63      1000
           5       0.72      0.54      0.61      1000
           6       0.59      0.91      0.71      1000
           7       0.81      0.71      0.76      1000
           8       0.70      0.92      0.79      1000
           9       0.90      0.77      0.83      1000

    accuracy                           0.70     10000
   macro avg       0.73      0.70      0.70     10000
weighted avg       0.73      0.70      0.70     10000

Accuracy Score: 0.7034
Root Mean Square Error: 2.2119448456053328
No description has been provided for this image
No description has been provided for this image
In [ ]:
data_df = pd.DataFrame(data)
data_df
Out[ ]:
model accuracy val_accuracy test_accuracy loss val_loss test_loss time
0 DNN 0.565 0.562 0.567 1.245 1.274 1.268 46.604
1 CNN_DO_MP_DO_16 0.634 0.610 0.619 1.056 1.110 1.108 60.615
2 CNN_DO_MP_DO_32 0.684 0.650 0.640 0.918 1.014 1.041 85.284
3 CNN_DO_MP_DO_64 0.726 0.662 0.656 0.792 0.968 1.000 82.357
4 CNN_DO_MP_DO_128 0.722 0.642 0.650 0.803 1.019 1.041 55.198
5 CNN_DO_MP_DO_256 0.746 0.646 0.633 0.728 1.034 1.080 106.222
6 CNN_DO_MP_DO_512 0.764 0.643 0.635 0.677 1.065 1.105 200.123
7 CNN_DO_MP_DO_32-2.2 0.662 0.632 0.628 0.978 1.077 1.080 100.190
8 CNN_DO_MP_DO_32-4.4 0.675 0.636 0.641 0.938 1.054 1.053 74.688
9 CNN_DO_MP_DO_32-5.5 0.674 0.639 0.641 0.943 1.043 1.058 68.455
10 CNN_DO_MP_DO_64-2.2 0.655 0.615 0.618 1.003 1.119 1.122 63.568
11 CNN_DO_MP_DO_64-4.4 0.691 0.621 0.615 0.894 1.081 1.113 58.112
12 CNN_DO_MP_DO_64-5.5 0.676 0.633 0.635 0.938 1.050 1.062 45.898
13 CNN_DO_MP_DO_128-2.2 0.715 0.627 0.627 0.814 1.060 1.074 79.575
14 CNN_DO_MP_DO_128-4.4 0.746 0.637 0.636 0.724 1.079 1.110 82.708
15 CNN_DO_MP_DO_128-5.5 0.721 0.632 0.642 0.807 1.068 1.075 73.661
16 CNN_DO_MP_DO_CNN_DO_MP_DO16-2.2 0.590 0.617 0.630 1.167 1.184 1.189 75.247
17 CNN_DO_MP_DO_CNN_DO_MP_DO_64_64 0.703 0.688 0.690 0.858 0.939 0.940 70.261
18 CNN_DO_MP_DO_CNN_DO_MP_DO_64_128 0.752 0.723 0.733 0.714 0.815 0.822 107.098
19 CNN_DO_MP_DO_CNN_DO_MP_DO_64_256 0.781 0.741 0.744 0.629 0.789 0.793 111.724
20 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.732 0.726 0.730 0.772 0.813 0.822 71.486
21 CNN_DO_MP_DO_CNN_DO_MP_DO_64_64 0.709 0.711 0.704 0.838 0.912 0.923 88.232
22 CNN_DO_MP_DO_CNN_DO_MP_DO_64_128 0.745 0.709 0.708 0.735 0.880 0.887 76.367
23 CNN_DO_MP_DO_CNN_DO_MP_DO_64_256 0.788 0.740 0.742 0.609 0.777 0.787 100.630
24 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.737 0.713 0.711 0.758 0.856 0.877 75.341
25 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.778 0.732 0.740 0.639 0.763 0.769 136.856
26 CNN_DO_MP_DO_X3_128_256_512 0.753 0.721 0.718 0.702 0.833 0.853 161.721
27 CNN_DO_MP_DO_X3_64_128_256 0.765 0.761 0.752 0.666 0.719 0.748 133.832
28 CNN_DO_MP_DO_X3_32_64_128 0.703 0.713 0.718 0.847 0.873 0.884 102.764
29 CNN_DO_MP_DO_X3_64_128_256-1 0.735 0.718 0.716 0.759 0.837 0.857 72.193
30 CNN_DO_MP_DO_X3_64_128_256-2 0.750 0.745 0.739 0.713 0.776 0.789 101.459
31 CNN_DO_MP_DO_X3_64_128_256-DNN 0.716 0.703 0.703 0.811 0.850 0.866 114.463
In [ ]:
for i in range(3):
  # Build Model 2-4: CNN and MaxPool with dropouts between and after
  name = f'CNN_DO_MP_DO_X3_64_128_256-Pat-3-{i}'
  k.clear_session()
  model = models.Sequential()
  model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1),activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Flatten())
  model.add(layers.Dense(units=10, activation=tf.nn.softmax))
  keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
  model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
  time_start = time.time()
  history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                      ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=3)])
  time_end = time.time()
  preds = model.predict(image_test_norm)
  test_pred = model.evaluate(image_test_norm, test_labels)

  history_dict = history.history
  history_df=pd.DataFrame(history_dict)
  plt.subplots(figsize=(16,12))
  plt.tight_layout()
  display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
  display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
  pred= model.predict(image_test_norm)
  pred=np.argmax(pred, axis=1)
  print_validation_report(test_labels, pred)
  plot_confusion_matrix(test_labels, pred)
  add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 32s 24ms/step - accuracy: 0.3265 - loss: 1.8252 - val_accuracy: 0.5456 - val_loss: 1.4853
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 18s 7ms/step - accuracy: 0.5434 - loss: 1.2788 - val_accuracy: 0.6184 - val_loss: 1.2192
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6099 - loss: 1.1130 - val_accuracy: 0.6550 - val_loss: 1.1546
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6432 - loss: 1.0176 - val_accuracy: 0.6656 - val_loss: 1.0567
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6689 - loss: 0.9475 - val_accuracy: 0.6882 - val_loss: 1.0203
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6809 - loss: 0.9108 - val_accuracy: 0.6886 - val_loss: 0.9760
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6942 - loss: 0.8733 - val_accuracy: 0.6992 - val_loss: 0.9488
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7109 - loss: 0.8291 - val_accuracy: 0.7272 - val_loss: 0.8831
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7180 - loss: 0.8169 - val_accuracy: 0.7086 - val_loss: 0.8900
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7238 - loss: 0.7892 - val_accuracy: 0.7234 - val_loss: 0.8617
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7297 - loss: 0.7720 - val_accuracy: 0.7202 - val_loss: 0.8336
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.7357 - loss: 0.8347
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.81      0.76      0.78      1000
           1       0.88      0.87      0.88      1000
           2       0.72      0.49      0.59      1000
           3       0.57      0.54      0.56      1000
           4       0.60      0.81      0.69      1000
           5       0.72      0.58      0.65      1000
           6       0.60      0.91      0.72      1000
           7       0.84      0.74      0.78      1000
           8       0.81      0.89      0.85      1000
           9       0.92      0.72      0.81      1000

    accuracy                           0.73     10000
   macro avg       0.75      0.73      0.73     10000
weighted avg       0.75      0.73      0.73     10000

Accuracy Score: 0.7313
Root Mean Square Error: 1.993589727100338

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 18s 15ms/step - accuracy: 0.3190 - loss: 1.8456 - val_accuracy: 0.5394 - val_loss: 1.4768
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.5372 - loss: 1.2956 - val_accuracy: 0.6254 - val_loss: 1.2632
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6077 - loss: 1.1196 - val_accuracy: 0.6536 - val_loss: 1.1471
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6416 - loss: 1.0268 - val_accuracy: 0.6716 - val_loss: 1.0740
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6618 - loss: 0.9644 - val_accuracy: 0.6836 - val_loss: 1.0278
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6792 - loss: 0.9152 - val_accuracy: 0.6728 - val_loss: 1.0023
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6925 - loss: 0.8783 - val_accuracy: 0.6920 - val_loss: 0.9195
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7040 - loss: 0.8503 - val_accuracy: 0.7270 - val_loss: 0.8619
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7137 - loss: 0.8241 - val_accuracy: 0.7028 - val_loss: 0.9029
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7167 - loss: 0.7998 - val_accuracy: 0.7160 - val_loss: 0.8757
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7254 - loss: 0.7862 - val_accuracy: 0.7144 - val_loss: 0.8716
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.7099 - loss: 0.8803
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.82      0.76      0.78      1000
           1       0.92      0.82      0.87      1000
           2       0.70      0.51      0.59      1000
           3       0.55      0.51      0.53      1000
           4       0.59      0.74      0.65      1000
           5       0.83      0.42      0.55      1000
           6       0.52      0.95      0.67      1000
           7       0.86      0.69      0.77      1000
           8       0.76      0.91      0.83      1000
           9       0.84      0.82      0.83      1000

    accuracy                           0.71     10000
   macro avg       0.74      0.71      0.71     10000
weighted avg       0.74      0.71      0.71     10000

Accuracy Score: 0.7106
Root Mean Square Error: 2.116719159454083

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 17s 14ms/step - accuracy: 0.3236 - loss: 1.8244 - val_accuracy: 0.5452 - val_loss: 1.4646
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 7ms/step - accuracy: 0.5489 - loss: 1.2717 - val_accuracy: 0.5994 - val_loss: 1.3000
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6084 - loss: 1.1187 - val_accuracy: 0.6270 - val_loss: 1.1636
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6391 - loss: 1.0263 - val_accuracy: 0.6900 - val_loss: 1.0586
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6671 - loss: 0.9550 - val_accuracy: 0.6938 - val_loss: 1.0261
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6851 - loss: 0.9013 - val_accuracy: 0.7058 - val_loss: 0.9347
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6960 - loss: 0.8743 - val_accuracy: 0.7200 - val_loss: 0.8955
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7087 - loss: 0.8344 - val_accuracy: 0.7352 - val_loss: 0.8849
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7177 - loss: 0.8099 - val_accuracy: 0.7322 - val_loss: 0.8682
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7214 - loss: 0.7940 - val_accuracy: 0.7268 - val_loss: 0.8537
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7349 - loss: 0.7623 - val_accuracy: 0.7384 - val_loss: 0.8457
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7404 - loss: 0.7435 - val_accuracy: 0.7070 - val_loss: 0.8796
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7428 - loss: 0.7270 - val_accuracy: 0.7584 - val_loss: 0.8012
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7480 - loss: 0.7229 - val_accuracy: 0.7520 - val_loss: 0.7785
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7560 - loss: 0.7026 - val_accuracy: 0.7696 - val_loss: 0.7566
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 7ms/step - accuracy: 0.7530 - loss: 0.7066 - val_accuracy: 0.7642 - val_loss: 0.7326
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7575 - loss: 0.6916 - val_accuracy: 0.7558 - val_loss: 0.7510
Epoch 18/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7592 - loss: 0.6799 - val_accuracy: 0.7586 - val_loss: 0.7659
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.7510 - loss: 0.7846
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.82      0.78      0.80      1000
           1       0.92      0.84      0.88      1000
           2       0.67      0.63      0.65      1000
           3       0.60      0.57      0.58      1000
           4       0.61      0.80      0.69      1000
           5       0.74      0.60      0.66      1000
           6       0.67      0.88      0.76      1000
           7       0.89      0.71      0.79      1000
           8       0.83      0.87      0.85      1000
           9       0.87      0.81      0.84      1000

    accuracy                           0.75     10000
   macro avg       0.76      0.75      0.75     10000
weighted avg       0.76      0.75      0.75     10000

Accuracy Score: 0.7486
Root Mean Square Error: 1.9880140844571499
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
In [ ]:
for i in range(3):
  # Build Model 2-4: CNN and MaxPool with dropouts between and after
  name = f'CNN_DO_MP_DO_X3_64_128_256-Pat-2-{i}'
  k.clear_session()
  model = models.Sequential()
  model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1),activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Flatten())
  model.add(layers.Dense(units=10, activation=tf.nn.softmax))
  keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
  model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
  time_start = time.time()
  history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                      ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=2)])
  time_end = time.time()
  preds = model.predict(image_test_norm)
  test_pred = model.evaluate(image_test_norm, test_labels)

  history_dict = history.history
  history_df=pd.DataFrame(history_dict)
  plt.subplots(figsize=(16,12))
  plt.tight_layout()
  display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
  display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
  pred= model.predict(image_test_norm)
  pred=np.argmax(pred, axis=1)
  print_validation_report(test_labels, pred)
  plot_confusion_matrix(test_labels, pred)
  add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 17s 15ms/step - accuracy: 0.3383 - loss: 1.8005 - val_accuracy: 0.5728 - val_loss: 1.4031
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 7ms/step - accuracy: 0.5574 - loss: 1.2574 - val_accuracy: 0.6096 - val_loss: 1.2575
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6089 - loss: 1.1148 - val_accuracy: 0.6592 - val_loss: 1.1421
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6471 - loss: 1.0145 - val_accuracy: 0.7006 - val_loss: 1.0201
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6691 - loss: 0.9474 - val_accuracy: 0.6498 - val_loss: 1.0551
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6902 - loss: 0.8934 - val_accuracy: 0.7154 - val_loss: 0.9913
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6994 - loss: 0.8586 - val_accuracy: 0.7054 - val_loss: 0.9069
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7091 - loss: 0.8306 - val_accuracy: 0.7164 - val_loss: 0.8988
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7182 - loss: 0.8039 - val_accuracy: 0.7364 - val_loss: 0.8526
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7270 - loss: 0.7837 - val_accuracy: 0.7336 - val_loss: 0.8486
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7356 - loss: 0.7611 - val_accuracy: 0.7374 - val_loss: 0.8222
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7370 - loss: 0.7544 - val_accuracy: 0.7442 - val_loss: 0.8146
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 7ms/step - accuracy: 0.7444 - loss: 0.7300 - val_accuracy: 0.7414 - val_loss: 0.8067
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7486 - loss: 0.7205 - val_accuracy: 0.7406 - val_loss: 0.7955
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.7400 - loss: 0.8178
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.81      0.78      0.79      1000
           1       0.93      0.77      0.84      1000
           2       0.77      0.52      0.62      1000
           3       0.61      0.52      0.56      1000
           4       0.55      0.83      0.66      1000
           5       0.67      0.65      0.66      1000
           6       0.63      0.91      0.74      1000
           7       0.87      0.68      0.77      1000
           8       0.80      0.89      0.84      1000
           9       0.88      0.78      0.83      1000

    accuracy                           0.73     10000
   macro avg       0.75      0.73      0.73     10000
weighted avg       0.75      0.73      0.73     10000

Accuracy Score: 0.7324
Root Mean Square Error: 2.0436976292984244

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 16s 15ms/step - accuracy: 0.3329 - loss: 1.8143 - val_accuracy: 0.5280 - val_loss: 1.4747
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.5409 - loss: 1.2889 - val_accuracy: 0.5850 - val_loss: 1.3343
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6025 - loss: 1.1322 - val_accuracy: 0.6304 - val_loss: 1.2018
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6362 - loss: 1.0440 - val_accuracy: 0.6784 - val_loss: 1.0791
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6541 - loss: 0.9845 - val_accuracy: 0.7004 - val_loss: 1.0153
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6817 - loss: 0.9118 - val_accuracy: 0.7106 - val_loss: 0.9834
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6957 - loss: 0.8760 - val_accuracy: 0.7154 - val_loss: 0.9237
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7088 - loss: 0.8401 - val_accuracy: 0.7252 - val_loss: 0.8850
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7126 - loss: 0.8141 - val_accuracy: 0.7326 - val_loss: 0.8660
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7234 - loss: 0.7947 - val_accuracy: 0.7316 - val_loss: 0.8715
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7262 - loss: 0.7811 - val_accuracy: 0.7330 - val_loss: 0.8350
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7324 - loss: 0.7717 - val_accuracy: 0.7324 - val_loss: 0.8270
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7430 - loss: 0.7404 - val_accuracy: 0.7348 - val_loss: 0.8211
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7508 - loss: 0.7141 - val_accuracy: 0.7464 - val_loss: 0.7990
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7526 - loss: 0.7140 - val_accuracy: 0.7400 - val_loss: 0.8199
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 7ms/step - accuracy: 0.7544 - loss: 0.7087 - val_accuracy: 0.7410 - val_loss: 0.7850
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - accuracy: 0.7315 - loss: 0.8067
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.84      0.76      0.79      1000
           1       0.91      0.86      0.89      1000
           2       0.72      0.54      0.62      1000
           3       0.56      0.55      0.56      1000
           4       0.58      0.81      0.68      1000
           5       0.76      0.54      0.63      1000
           6       0.58      0.92      0.71      1000
           7       0.87      0.71      0.78      1000
           8       0.83      0.87      0.85      1000
           9       0.89      0.79      0.84      1000

    accuracy                           0.73     10000
   macro avg       0.76      0.73      0.73     10000
weighted avg       0.76      0.73      0.73     10000

Accuracy Score: 0.7349
Root Mean Square Error: 1.9666723163760658

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 17s 14ms/step - accuracy: 0.3263 - loss: 1.8125 - val_accuracy: 0.5112 - val_loss: 1.4722
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 7ms/step - accuracy: 0.5433 - loss: 1.2738 - val_accuracy: 0.6184 - val_loss: 1.2663
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6037 - loss: 1.1140 - val_accuracy: 0.6486 - val_loss: 1.1542
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6421 - loss: 1.0190 - val_accuracy: 0.6600 - val_loss: 1.0827
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6639 - loss: 0.9693 - val_accuracy: 0.6782 - val_loss: 1.0224
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6881 - loss: 0.8982 - val_accuracy: 0.7050 - val_loss: 0.9507
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6980 - loss: 0.8725 - val_accuracy: 0.7172 - val_loss: 0.9161
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7061 - loss: 0.8456 - val_accuracy: 0.6622 - val_loss: 0.9808
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7184 - loss: 0.8028 - val_accuracy: 0.7026 - val_loss: 0.8990
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.6998 - loss: 0.9190
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.84      0.69      0.76      1000
           1       0.92      0.81      0.86      1000
           2       0.65      0.54      0.59      1000
           3       0.57      0.47      0.51      1000
           4       0.55      0.76      0.64      1000
           5       0.66      0.57      0.61      1000
           6       0.55      0.94      0.69      1000
           7       0.90      0.60      0.72      1000
           8       0.77      0.87      0.82      1000
           9       0.89      0.76      0.82      1000

    accuracy                           0.70     10000
   macro avg       0.73      0.70      0.70     10000
weighted avg       0.73      0.70      0.70     10000

Accuracy Score: 0.702
Root Mean Square Error: 2.0995237555217137
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
In [ ]:
for i in range(3):
  # Build Model 2-4: CNN and MaxPool with dropouts between and after
  name = f'CNN_DO_MP_DO_X3_64_128_256-Pat-1-{i}'
  k.clear_session()
  model = models.Sequential()
  model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1),activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Flatten())
  model.add(layers.Dense(units=10, activation=tf.nn.softmax))
  keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
  model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
  time_start = time.time()
  history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                      ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=1)])
  time_end = time.time()
  preds = model.predict(image_test_norm)
  test_pred = model.evaluate(image_test_norm, test_labels)

  history_dict = history.history
  history_df=pd.DataFrame(history_dict)
  plt.subplots(figsize=(16,12))
  plt.tight_layout()
  display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
  display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
  pred= model.predict(image_test_norm)
  pred=np.argmax(pred, axis=1)
  print_validation_report(test_labels, pred)
  plot_confusion_matrix(test_labels, pred)
  add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 15s 13ms/step - accuracy: 0.3270 - loss: 1.8141 - val_accuracy: 0.5348 - val_loss: 1.4791
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5310 - loss: 1.3056 - val_accuracy: 0.6092 - val_loss: 1.2793
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.5975 - loss: 1.1469 - val_accuracy: 0.6370 - val_loss: 1.1749
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6362 - loss: 1.0363 - val_accuracy: 0.6780 - val_loss: 1.0594
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6608 - loss: 0.9642 - val_accuracy: 0.6888 - val_loss: 1.0420
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6752 - loss: 0.9287 - val_accuracy: 0.6878 - val_loss: 0.9944
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 5ms/step - accuracy: 0.6918 - loss: 1.0046
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.78      0.74      0.76      1000
           1       0.87      0.81      0.84      1000
           2       0.62      0.55      0.58      1000
           3       0.46      0.58      0.51      1000
           4       0.52      0.78      0.63      1000
           5       0.56      0.65      0.60      1000
           6       0.81      0.68      0.73      1000
           7       0.86      0.60      0.70      1000
           8       0.79      0.83      0.81      1000
           9       0.88      0.68      0.77      1000

    accuracy                           0.69     10000
   macro avg       0.72      0.69      0.69     10000
weighted avg       0.72      0.69      0.69     10000

Accuracy Score: 0.6881
Root Mean Square Error: 2.1743734729802053

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 16s 14ms/step - accuracy: 0.3216 - loss: 1.8306 - val_accuracy: 0.4618 - val_loss: 1.6185
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 6ms/step - accuracy: 0.5421 - loss: 1.2891 - val_accuracy: 0.6244 - val_loss: 1.2552
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6093 - loss: 1.1033 - val_accuracy: 0.6464 - val_loss: 1.1279
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6449 - loss: 1.0141 - val_accuracy: 0.6866 - val_loss: 1.0180
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6685 - loss: 0.9418 - val_accuracy: 0.6818 - val_loss: 0.9982
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 5ms/step - accuracy: 0.6916 - loss: 1.0109
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.83      0.58      0.68      1000
           1       0.95      0.71      0.81      1000
           2       0.63      0.51      0.56      1000
           3       0.47      0.67      0.55      1000
           4       0.54      0.77      0.63      1000
           5       0.66      0.55      0.60      1000
           6       0.80      0.72      0.76      1000
           7       0.81      0.66      0.73      1000
           8       0.64      0.93      0.76      1000
           9       0.84      0.75      0.79      1000

    accuracy                           0.69     10000
   macro avg       0.72      0.69      0.69     10000
weighted avg       0.72      0.69      0.69     10000

Accuracy Score: 0.6857
Root Mean Square Error: 2.305775357661713

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 16s 14ms/step - accuracy: 0.3284 - loss: 1.8136 - val_accuracy: 0.5510 - val_loss: 1.4374
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 7ms/step - accuracy: 0.5519 - loss: 1.2567 - val_accuracy: 0.6390 - val_loss: 1.2332
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 8ms/step - accuracy: 0.6172 - loss: 1.0958 - val_accuracy: 0.6776 - val_loss: 1.1067
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.6507 - loss: 1.0020 - val_accuracy: 0.6996 - val_loss: 1.0651
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.6670 - loss: 0.9535 - val_accuracy: 0.6980 - val_loss: 1.0101
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 5ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 6ms/step - accuracy: 0.6986 - loss: 1.0182
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.81      0.71      0.76      1000
           1       0.89      0.82      0.85      1000
           2       0.71      0.42      0.53      1000
           3       0.49      0.52      0.50      1000
           4       0.68      0.57      0.62      1000
           5       0.52      0.73      0.61      1000
           6       0.60      0.90      0.72      1000
           7       0.78      0.71      0.74      1000
           8       0.82      0.83      0.83      1000
           9       0.86      0.76      0.81      1000

    accuracy                           0.70     10000
   macro avg       0.72      0.70      0.70     10000
weighted avg       0.72      0.70      0.70     10000

Accuracy Score: 0.6978
Root Mean Square Error: 2.129248693788492
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
In [ ]:
data_df = pd.DataFrame(data)
data_df
Out[ ]:
model accuracy val_accuracy test_accuracy loss val_loss test_loss time
0 DNN 0.565 0.562 0.567 1.245 1.274 1.268 46.604
1 CNN_DO_MP_DO_16 0.634 0.610 0.619 1.056 1.110 1.108 60.615
2 CNN_DO_MP_DO_32 0.684 0.650 0.640 0.918 1.014 1.041 85.284
3 CNN_DO_MP_DO_64 0.726 0.662 0.656 0.792 0.968 1.000 82.357
4 CNN_DO_MP_DO_128 0.722 0.642 0.650 0.803 1.019 1.041 55.198
5 CNN_DO_MP_DO_256 0.746 0.646 0.633 0.728 1.034 1.080 106.222
6 CNN_DO_MP_DO_512 0.764 0.643 0.635 0.677 1.065 1.105 200.123
7 CNN_DO_MP_DO_32-2.2 0.662 0.632 0.628 0.978 1.077 1.080 100.190
8 CNN_DO_MP_DO_32-4.4 0.675 0.636 0.641 0.938 1.054 1.053 74.688
9 CNN_DO_MP_DO_32-5.5 0.674 0.639 0.641 0.943 1.043 1.058 68.455
10 CNN_DO_MP_DO_64-2.2 0.655 0.615 0.618 1.003 1.119 1.122 63.568
11 CNN_DO_MP_DO_64-4.4 0.691 0.621 0.615 0.894 1.081 1.113 58.112
12 CNN_DO_MP_DO_64-5.5 0.676 0.633 0.635 0.938 1.050 1.062 45.898
13 CNN_DO_MP_DO_128-2.2 0.715 0.627 0.627 0.814 1.060 1.074 79.575
14 CNN_DO_MP_DO_128-4.4 0.746 0.637 0.636 0.724 1.079 1.110 82.708
15 CNN_DO_MP_DO_128-5.5 0.721 0.632 0.642 0.807 1.068 1.075 73.661
16 CNN_DO_MP_DO_CNN_DO_MP_DO16-2.2 0.590 0.617 0.630 1.167 1.184 1.189 75.247
17 CNN_DO_MP_DO_CNN_DO_MP_DO_64_64 0.703 0.688 0.690 0.858 0.939 0.940 70.261
18 CNN_DO_MP_DO_CNN_DO_MP_DO_64_128 0.752 0.723 0.733 0.714 0.815 0.822 107.098
19 CNN_DO_MP_DO_CNN_DO_MP_DO_64_256 0.781 0.741 0.744 0.629 0.789 0.793 111.724
20 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.732 0.726 0.730 0.772 0.813 0.822 71.486
21 CNN_DO_MP_DO_CNN_DO_MP_DO_64_64 0.709 0.711 0.704 0.838 0.912 0.923 88.232
22 CNN_DO_MP_DO_CNN_DO_MP_DO_64_128 0.745 0.709 0.708 0.735 0.880 0.887 76.367
23 CNN_DO_MP_DO_CNN_DO_MP_DO_64_256 0.788 0.740 0.742 0.609 0.777 0.787 100.630
24 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.737 0.713 0.711 0.758 0.856 0.877 75.341
25 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.778 0.732 0.740 0.639 0.763 0.769 136.856
26 CNN_DO_MP_DO_X3_128_256_512 0.753 0.721 0.718 0.702 0.833 0.853 161.721
27 CNN_DO_MP_DO_X3_64_128_256 0.765 0.761 0.752 0.666 0.719 0.748 133.832
28 CNN_DO_MP_DO_X3_32_64_128 0.703 0.713 0.718 0.847 0.873 0.884 102.764
29 CNN_DO_MP_DO_X3_64_128_256-1 0.735 0.718 0.716 0.759 0.837 0.857 72.193
30 CNN_DO_MP_DO_X3_64_128_256-2 0.750 0.745 0.739 0.713 0.776 0.789 101.459
31 CNN_DO_MP_DO_X3_64_128_256-DNN 0.716 0.703 0.703 0.811 0.850 0.866 114.463
32 CNN_DO_MP_DO_X3_64_128_256-Pat-3-0 0.731 0.720 0.731 0.773 0.834 0.840 104.831
33 CNN_DO_MP_DO_X3_64_128_256-Pat-3-1 0.724 0.714 0.711 0.789 0.872 0.883 68.565
34 CNN_DO_MP_DO_X3_64_128_256-Pat-3-2 0.761 0.759 0.749 0.683 0.766 0.788 107.973
35 CNN_DO_MP_DO_X3_64_128_256-Pat-2-0 0.748 0.741 0.732 0.724 0.795 0.819 87.255
36 CNN_DO_MP_DO_X3_64_128_256-Pat-2-1 0.750 0.741 0.735 0.716 0.785 0.802 91.753
37 CNN_DO_MP_DO_X3_64_128_256-Pat-2-2 0.719 0.703 0.702 0.806 0.899 0.914 62.646
38 CNN_DO_MP_DO_X3_64_128_256-Pat-1-0 0.676 0.688 0.688 0.929 0.994 1.005 41.329
39 CNN_DO_MP_DO_X3_64_128_256-Pat-1-1 0.673 0.682 0.686 0.933 0.998 1.012 42.944
40 CNN_DO_MP_DO_X3_64_128_256-Pat-1-2 0.674 0.698 0.698 0.938 1.010 1.021 50.324
In [ ]:
for i in range(3):
  # Build Model 2-4: CNN and MaxPool with dropouts between and after
  name = f'CNN_DO_MP_DO_X3_64_128_256-Pat-4-{i}'
  k.clear_session()
  model = models.Sequential()
  model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1),activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Flatten())
  model.add(layers.Dense(units=10, activation=tf.nn.softmax))
  keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
  model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
  time_start = time.time()
  history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                      ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=4)])
  time_end = time.time()
  preds = model.predict(image_test_norm)
  test_pred = model.evaluate(image_test_norm, test_labels)

  history_dict = history.history
  history_df=pd.DataFrame(history_dict)
  plt.subplots(figsize=(16,12))
  plt.tight_layout()
  display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
  display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
  pred= model.predict(image_test_norm)
  pred=np.argmax(pred, axis=1)
  print_validation_report(test_labels, pred)
  plot_confusion_matrix(test_labels, pred)
  add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 19s 14ms/step - accuracy: 0.3183 - loss: 1.8414 - val_accuracy: 0.5372 - val_loss: 1.4382
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5495 - loss: 1.2722 - val_accuracy: 0.5974 - val_loss: 1.3085
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5995 - loss: 1.1256 - val_accuracy: 0.6524 - val_loss: 1.1627
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6412 - loss: 1.0276 - val_accuracy: 0.6690 - val_loss: 1.0488
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6628 - loss: 0.9685 - val_accuracy: 0.6956 - val_loss: 0.9989
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6722 - loss: 0.9255 - val_accuracy: 0.7068 - val_loss: 0.9435
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6906 - loss: 0.8782 - val_accuracy: 0.7106 - val_loss: 0.9220
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7004 - loss: 0.8510 - val_accuracy: 0.7204 - val_loss: 0.8804
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7132 - loss: 0.8254 - val_accuracy: 0.7268 - val_loss: 0.8551
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7144 - loss: 0.8193 - val_accuracy: 0.6730 - val_loss: 0.9627
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7215 - loss: 0.7977 - val_accuracy: 0.7040 - val_loss: 0.8969
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7291 - loss: 0.7782 - val_accuracy: 0.7426 - val_loss: 0.8388
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7319 - loss: 0.7663 - val_accuracy: 0.7282 - val_loss: 0.8185
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7378 - loss: 0.7437 - val_accuracy: 0.7440 - val_loss: 0.8192
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7456 - loss: 0.7267 - val_accuracy: 0.7356 - val_loss: 0.8055
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7526 - loss: 0.7079 - val_accuracy: 0.7672 - val_loss: 0.7549
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7527 - loss: 0.7064 - val_accuracy: 0.7510 - val_loss: 0.7658
Epoch 18/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7558 - loss: 0.6973 - val_accuracy: 0.7178 - val_loss: 0.8345
Epoch 19/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7615 - loss: 0.6821 - val_accuracy: 0.7530 - val_loss: 0.7724
Epoch 20/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7628 - loss: 0.6823 - val_accuracy: 0.7514 - val_loss: 0.7643
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.7505 - loss: 0.7771
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.80      0.79      0.79      1000
           1       0.91      0.86      0.88      1000
           2       0.73      0.59      0.65      1000
           3       0.66      0.50      0.57      1000
           4       0.58      0.85      0.69      1000
           5       0.72      0.62      0.67      1000
           6       0.64      0.92      0.75      1000
           7       0.91      0.67      0.77      1000
           8       0.80      0.90      0.85      1000
           9       0.90      0.77      0.83      1000

    accuracy                           0.75     10000
   macro avg       0.76      0.75      0.75     10000
weighted avg       0.76      0.75      0.75     10000

Accuracy Score: 0.7471
Root Mean Square Error: 1.9793433254491248

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 16s 14ms/step - accuracy: 0.3113 - loss: 1.8545 - val_accuracy: 0.5356 - val_loss: 1.4574
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 6ms/step - accuracy: 0.5311 - loss: 1.3102 - val_accuracy: 0.5748 - val_loss: 1.3113
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5953 - loss: 1.1580 - val_accuracy: 0.6484 - val_loss: 1.1747
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6316 - loss: 1.0589 - val_accuracy: 0.6622 - val_loss: 1.1019
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 7ms/step - accuracy: 0.6583 - loss: 0.9758 - val_accuracy: 0.6848 - val_loss: 1.0346
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.6755 - loss: 0.9318 - val_accuracy: 0.6936 - val_loss: 1.0031
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6844 - loss: 0.9009 - val_accuracy: 0.7026 - val_loss: 0.9818
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7005 - loss: 0.8692 - val_accuracy: 0.6882 - val_loss: 0.9608
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7089 - loss: 0.8375 - val_accuracy: 0.7214 - val_loss: 0.8772
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7157 - loss: 0.8166 - val_accuracy: 0.7354 - val_loss: 0.8346
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7189 - loss: 0.8001 - val_accuracy: 0.7270 - val_loss: 0.8548
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7283 - loss: 0.7742 - val_accuracy: 0.7270 - val_loss: 0.8576
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7313 - loss: 0.7659 - val_accuracy: 0.7216 - val_loss: 0.8352
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.7396 - loss: 0.7419 - val_accuracy: 0.7438 - val_loss: 0.7804
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7431 - loss: 0.7303 - val_accuracy: 0.7140 - val_loss: 0.8501
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7466 - loss: 0.7182 - val_accuracy: 0.7546 - val_loss: 0.7979
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7514 - loss: 0.7091 - val_accuracy: 0.7376 - val_loss: 0.7990
Epoch 18/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 7ms/step - accuracy: 0.7513 - loss: 0.7143 - val_accuracy: 0.7262 - val_loss: 0.7975
Epoch 19/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7544 - loss: 0.7123 - val_accuracy: 0.7552 - val_loss: 0.7748
Epoch 20/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7574 - loss: 0.6942 - val_accuracy: 0.6832 - val_loss: 0.9027
Epoch 21/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7610 - loss: 0.6847 - val_accuracy: 0.7426 - val_loss: 0.7714
Epoch 22/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7678 - loss: 0.6683 - val_accuracy: 0.7636 - val_loss: 0.7585
Epoch 23/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7639 - loss: 0.6716 - val_accuracy: 0.7460 - val_loss: 0.7521
Epoch 24/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7670 - loss: 0.6637 - val_accuracy: 0.7650 - val_loss: 0.7274
Epoch 25/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7662 - loss: 0.6586 - val_accuracy: 0.7758 - val_loss: 0.7080
Epoch 26/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7736 - loss: 0.6459 - val_accuracy: 0.7464 - val_loss: 0.7421
Epoch 27/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7692 - loss: 0.6516 - val_accuracy: 0.7294 - val_loss: 0.7768
Epoch 28/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7797 - loss: 0.6280 - val_accuracy: 0.7636 - val_loss: 0.7143
Epoch 29/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7805 - loss: 0.6271 - val_accuracy: 0.7666 - val_loss: 0.7057
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.7656 - loss: 0.7070
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.83      0.79      0.81      1000
           1       0.93      0.84      0.88      1000
           2       0.75      0.60      0.67      1000
           3       0.64      0.54      0.59      1000
           4       0.63      0.85      0.72      1000
           5       0.67      0.72      0.69      1000
           6       0.73      0.89      0.80      1000
           7       0.91      0.73      0.81      1000
           8       0.79      0.91      0.85      1000
           9       0.90      0.81      0.85      1000

    accuracy                           0.77     10000
   macro avg       0.78      0.77      0.77     10000
weighted avg       0.78      0.77      0.77     10000

Accuracy Score: 0.7682
Root Mean Square Error: 1.8783769589728256

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 17s 16ms/step - accuracy: 0.3147 - loss: 1.8437 - val_accuracy: 0.5312 - val_loss: 1.4586
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 7ms/step - accuracy: 0.5431 - loss: 1.2845 - val_accuracy: 0.6068 - val_loss: 1.2894
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6077 - loss: 1.1150 - val_accuracy: 0.6614 - val_loss: 1.1440
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6348 - loss: 1.0398 - val_accuracy: 0.6920 - val_loss: 1.0418
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6633 - loss: 0.9673 - val_accuracy: 0.6538 - val_loss: 1.0480
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6828 - loss: 0.9099 - val_accuracy: 0.6926 - val_loss: 0.9787
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6949 - loss: 0.8675 - val_accuracy: 0.6980 - val_loss: 0.9491
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7100 - loss: 0.8354 - val_accuracy: 0.6542 - val_loss: 1.0866
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7138 - loss: 0.8204 - val_accuracy: 0.7344 - val_loss: 0.8423
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7235 - loss: 0.7907 - val_accuracy: 0.7320 - val_loss: 0.8580
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7326 - loss: 0.7696 - val_accuracy: 0.6956 - val_loss: 0.8953
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7361 - loss: 0.7561 - val_accuracy: 0.7072 - val_loss: 0.8661
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7368 - loss: 0.7456 - val_accuracy: 0.6956 - val_loss: 0.8905
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.6900 - loss: 0.9114
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.84      0.65      0.73      1000
           1       0.96      0.73      0.83      1000
           2       0.66      0.55      0.60      1000
           3       0.53      0.51      0.52      1000
           4       0.60      0.70      0.65      1000
           5       0.85      0.39      0.53      1000
           6       0.49      0.95      0.65      1000
           7       0.90      0.65      0.75      1000
           8       0.67      0.92      0.78      1000
           9       0.79      0.82      0.81      1000

    accuracy                           0.69     10000
   macro avg       0.73      0.69      0.69     10000
weighted avg       0.73      0.69      0.69     10000

Accuracy Score: 0.6871
Root Mean Square Error: 2.283527972239447
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
In [ ]:
for i in range(3):
  # Build Model 2-4: CNN and MaxPool with dropouts between and after
  name = f'CNN_DO_MP_DO_X3_64_128_256-Pat-5-{i}'
  k.clear_session()
  model = models.Sequential()
  model.add(layers.Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), input_shape=((32,32,3,)), activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1),activation=tf.nn.relu))
  model.add(layers.Dropout(0.25))
  model.add(layers.MaxPool2D((2, 2),strides=2))
  model.add(layers.Dropout(0.25))
  model.add(layers.Flatten())
  model.add(layers.Dense(units=10, activation=tf.nn.softmax))
  keras.utils.plot_model(model, f"CIFAR10_{name}.png", show_shapes=True)
  model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False), metrics=['accuracy'])
  time_start = time.time()
  history = model.fit(image_train_norm, label_train_split, epochs=200, batch_size=64, validation_data=(image_val_norm, label_val_split), callbacks=[tf.keras.callbacks.ModelCheckpoint(f"{name}_model.keras",save_best_only=True,save_weights_only=False)
                      ,tf.keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=5)])
  time_end = time.time()
  preds = model.predict(image_test_norm)
  test_pred = model.evaluate(image_test_norm, test_labels)

  history_dict = history.history
  history_df=pd.DataFrame(history_dict)
  plt.subplots(figsize=(16,12))
  plt.tight_layout()
  display_training_curves(history_df['accuracy'], history_df['val_accuracy'], 'accuracy', 211)
  display_training_curves(history_df['loss'], history_df['val_loss'], 'loss', 212)
  pred= model.predict(image_test_norm)
  pred=np.argmax(pred, axis=1)
  print_validation_report(test_labels, pred)
  plot_confusion_matrix(test_labels, pred)
  add_to_data(data, name, history, test_pred)

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 15s 14ms/step - accuracy: 0.3266 - loss: 1.8180 - val_accuracy: 0.5104 - val_loss: 1.5165
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.5458 - loss: 1.2759 - val_accuracy: 0.6236 - val_loss: 1.2663
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6090 - loss: 1.1290 - val_accuracy: 0.6486 - val_loss: 1.1534
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6438 - loss: 1.0277 - val_accuracy: 0.6694 - val_loss: 1.0873
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6595 - loss: 0.9691 - val_accuracy: 0.6786 - val_loss: 1.0048
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6822 - loss: 0.9183 - val_accuracy: 0.7124 - val_loss: 0.9573
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6945 - loss: 0.8762 - val_accuracy: 0.6818 - val_loss: 0.9901
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7036 - loss: 0.8516 - val_accuracy: 0.6812 - val_loss: 0.9579
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7165 - loss: 0.8132 - val_accuracy: 0.6930 - val_loss: 0.9148
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7219 - loss: 0.8021 - val_accuracy: 0.7268 - val_loss: 0.8803
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7289 - loss: 0.7822 - val_accuracy: 0.7190 - val_loss: 0.8505
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7336 - loss: 0.7626 - val_accuracy: 0.7630 - val_loss: 0.7760
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7403 - loss: 0.7352 - val_accuracy: 0.7320 - val_loss: 0.8164
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 6ms/step - accuracy: 0.7434 - loss: 0.7337 - val_accuracy: 0.7468 - val_loss: 0.7918
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7483 - loss: 0.7216 - val_accuracy: 0.7460 - val_loss: 0.7902
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7508 - loss: 0.7092 - val_accuracy: 0.7310 - val_loss: 0.8168
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7570 - loss: 0.6955 - val_accuracy: 0.7398 - val_loss: 0.7869
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.7224 - loss: 0.8187
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.87      0.62      0.72      1000
           1       0.93      0.80      0.86      1000
           2       0.60      0.66      0.63      1000
           3       0.56      0.58      0.57      1000
           4       0.65      0.73      0.69      1000
           5       0.68      0.61      0.64      1000
           6       0.70      0.89      0.78      1000
           7       0.93      0.61      0.74      1000
           8       0.65      0.94      0.77      1000
           9       0.87      0.78      0.82      1000

    accuracy                           0.72     10000
   macro avg       0.74      0.72      0.72     10000
weighted avg       0.74      0.72      0.72     10000

Accuracy Score: 0.722
Root Mean Square Error: 2.1975213309544914

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 17s 15ms/step - accuracy: 0.3290 - loss: 1.7980 - val_accuracy: 0.5466 - val_loss: 1.4535
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 7ms/step - accuracy: 0.5548 - loss: 1.2459 - val_accuracy: 0.6164 - val_loss: 1.2509
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6161 - loss: 1.0973 - val_accuracy: 0.6446 - val_loss: 1.1623
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6513 - loss: 1.0017 - val_accuracy: 0.6382 - val_loss: 1.1363
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6686 - loss: 0.9507 - val_accuracy: 0.6708 - val_loss: 1.0397
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6887 - loss: 0.8987 - val_accuracy: 0.6834 - val_loss: 0.9740
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7007 - loss: 0.8582 - val_accuracy: 0.7094 - val_loss: 0.9391
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 7ms/step - accuracy: 0.7114 - loss: 0.8327 - val_accuracy: 0.7156 - val_loss: 0.9162
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7196 - loss: 0.7992 - val_accuracy: 0.7146 - val_loss: 0.8632
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7245 - loss: 0.7878 - val_accuracy: 0.7350 - val_loss: 0.8272
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7348 - loss: 0.7614 - val_accuracy: 0.7478 - val_loss: 0.7983
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7385 - loss: 0.7478 - val_accuracy: 0.7290 - val_loss: 0.8185
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 8ms/step - accuracy: 0.7429 - loss: 0.7326 - val_accuracy: 0.7398 - val_loss: 0.7977
Epoch 14/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 6ms/step - accuracy: 0.7507 - loss: 0.7111 - val_accuracy: 0.7356 - val_loss: 0.8049
Epoch 15/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7509 - loss: 0.7095 - val_accuracy: 0.7476 - val_loss: 0.7786
Epoch 16/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7587 - loss: 0.6885 - val_accuracy: 0.7544 - val_loss: 0.7596
Epoch 17/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7577 - loss: 0.6877 - val_accuracy: 0.7418 - val_loss: 0.7825
Epoch 18/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7711 - loss: 0.6641 - val_accuracy: 0.7324 - val_loss: 0.7786
Epoch 19/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7678 - loss: 0.6631 - val_accuracy: 0.7658 - val_loss: 0.7338
Epoch 20/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7713 - loss: 0.6521 - val_accuracy: 0.7502 - val_loss: 0.7357
Epoch 21/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7720 - loss: 0.6469 - val_accuracy: 0.7632 - val_loss: 0.7389
Epoch 22/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7766 - loss: 0.6363 - val_accuracy: 0.7544 - val_loss: 0.7339
Epoch 23/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7757 - loss: 0.6424 - val_accuracy: 0.7518 - val_loss: 0.7479
Epoch 24/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7807 - loss: 0.6360 - val_accuracy: 0.7564 - val_loss: 0.7170
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 5ms/step - accuracy: 0.7577 - loss: 0.7330
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.85      0.74      0.79      1000
           1       0.91      0.87      0.89      1000
           2       0.78      0.55      0.65      1000
           3       0.73      0.41      0.52      1000
           4       0.56      0.88      0.68      1000
           5       0.69      0.70      0.70      1000
           6       0.70      0.91      0.79      1000
           7       0.85      0.77      0.81      1000
           8       0.76      0.93      0.84      1000
           9       0.89      0.80      0.84      1000

    accuracy                           0.76     10000
   macro avg       0.77      0.76      0.75     10000
weighted avg       0.77      0.76      0.75     10000

Accuracy Score: 0.7555
Root Mean Square Error: 1.9526392395934278

/usr/local/lib/python3.11/dist-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
  super().__init__(activity_regularizer=activity_regularizer, **kwargs)

Epoch 1/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 16s 14ms/step - accuracy: 0.3306 - loss: 1.8029 - val_accuracy: 0.5406 - val_loss: 1.5078
Epoch 2/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 11s 7ms/step - accuracy: 0.5443 - loss: 1.2737 - val_accuracy: 0.5758 - val_loss: 1.3111
Epoch 3/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6092 - loss: 1.1121 - val_accuracy: 0.6606 - val_loss: 1.1303
Epoch 4/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6365 - loss: 1.0338 - val_accuracy: 0.6762 - val_loss: 1.0696
Epoch 5/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.6662 - loss: 0.9641 - val_accuracy: 0.6840 - val_loss: 1.0398
Epoch 6/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 7ms/step - accuracy: 0.6763 - loss: 0.9239 - val_accuracy: 0.7022 - val_loss: 0.9490
Epoch 7/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.6945 - loss: 0.8738 - val_accuracy: 0.7190 - val_loss: 0.9297
Epoch 8/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7052 - loss: 0.8390 - val_accuracy: 0.7274 - val_loss: 0.9033
Epoch 9/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7167 - loss: 0.8122 - val_accuracy: 0.7170 - val_loss: 0.8696
Epoch 10/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7179 - loss: 0.8012 - val_accuracy: 0.7134 - val_loss: 0.8852
Epoch 11/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 7ms/step - accuracy: 0.7287 - loss: 0.7757 - val_accuracy: 0.7040 - val_loss: 0.9116
Epoch 12/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7370 - loss: 0.7576 - val_accuracy: 0.7232 - val_loss: 0.8135
Epoch 13/200
704/704 ━━━━━━━━━━━━━━━━━━━━ 5s 6ms/step - accuracy: 0.7389 - loss: 0.7461 - val_accuracy: 0.7222 - val_loss: 0.8311
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 4ms/step - accuracy: 0.7212 - loss: 0.8508
313/313 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Classification Report
              precision    recall  f1-score   support

           0       0.78      0.80      0.79      1000
           1       0.86      0.89      0.88      1000
           2       0.62      0.63      0.62      1000
           3       0.58      0.48      0.53      1000
           4       0.69      0.68      0.68      1000
           5       0.77      0.47      0.58      1000
           6       0.52      0.94      0.67      1000
           7       0.85      0.72      0.78      1000
           8       0.82      0.87      0.84      1000
           9       0.92      0.72      0.81      1000

    accuracy                           0.72     10000
   macro avg       0.74      0.72      0.72     10000
weighted avg       0.74      0.72      0.72     10000

Accuracy Score: 0.7203
Root Mean Square Error: 2.0924148728203975
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
In [ ]:
data_df = pd.DataFrame(data)
data_df
Out[ ]:
model accuracy val_accuracy test_accuracy loss val_loss test_loss time
0 DNN 0.565 0.562 0.567 1.245 1.274 1.268 46.604
1 CNN_DO_MP_DO_16 0.634 0.610 0.619 1.056 1.110 1.108 60.615
2 CNN_DO_MP_DO_32 0.684 0.650 0.640 0.918 1.014 1.041 85.284
3 CNN_DO_MP_DO_64 0.726 0.662 0.656 0.792 0.968 1.000 82.357
4 CNN_DO_MP_DO_128 0.722 0.642 0.650 0.803 1.019 1.041 55.198
5 CNN_DO_MP_DO_256 0.746 0.646 0.633 0.728 1.034 1.080 106.222
6 CNN_DO_MP_DO_512 0.764 0.643 0.635 0.677 1.065 1.105 200.123
7 CNN_DO_MP_DO_32-2.2 0.662 0.632 0.628 0.978 1.077 1.080 100.190
8 CNN_DO_MP_DO_32-4.4 0.675 0.636 0.641 0.938 1.054 1.053 74.688
9 CNN_DO_MP_DO_32-5.5 0.674 0.639 0.641 0.943 1.043 1.058 68.455
10 CNN_DO_MP_DO_64-2.2 0.655 0.615 0.618 1.003 1.119 1.122 63.568
11 CNN_DO_MP_DO_64-4.4 0.691 0.621 0.615 0.894 1.081 1.113 58.112
12 CNN_DO_MP_DO_64-5.5 0.676 0.633 0.635 0.938 1.050 1.062 45.898
13 CNN_DO_MP_DO_128-2.2 0.715 0.627 0.627 0.814 1.060 1.074 79.575
14 CNN_DO_MP_DO_128-4.4 0.746 0.637 0.636 0.724 1.079 1.110 82.708
15 CNN_DO_MP_DO_128-5.5 0.721 0.632 0.642 0.807 1.068 1.075 73.661
16 CNN_DO_MP_DO_CNN_DO_MP_DO16-2.2 0.590 0.617 0.630 1.167 1.184 1.189 75.247
17 CNN_DO_MP_DO_CNN_DO_MP_DO_64_64 0.703 0.688 0.690 0.858 0.939 0.940 70.261
18 CNN_DO_MP_DO_CNN_DO_MP_DO_64_128 0.752 0.723 0.733 0.714 0.815 0.822 107.098
19 CNN_DO_MP_DO_CNN_DO_MP_DO_64_256 0.781 0.741 0.744 0.629 0.789 0.793 111.724
20 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.732 0.726 0.730 0.772 0.813 0.822 71.486
21 CNN_DO_MP_DO_CNN_DO_MP_DO_64_64 0.709 0.711 0.704 0.838 0.912 0.923 88.232
22 CNN_DO_MP_DO_CNN_DO_MP_DO_64_128 0.745 0.709 0.708 0.735 0.880 0.887 76.367
23 CNN_DO_MP_DO_CNN_DO_MP_DO_64_256 0.788 0.740 0.742 0.609 0.777 0.787 100.630
24 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.737 0.713 0.711 0.758 0.856 0.877 75.341
25 CNN_DO_MP_DO_CNN_DO_MP_DO_CNN_DO_MP_DO_64_128_256 0.778 0.732 0.740 0.639 0.763 0.769 136.856
26 CNN_DO_MP_DO_X3_128_256_512 0.753 0.721 0.718 0.702 0.833 0.853 161.721
27 CNN_DO_MP_DO_X3_64_128_256 0.765 0.761 0.752 0.666 0.719 0.748 133.832
28 CNN_DO_MP_DO_X3_32_64_128 0.703 0.713 0.718 0.847 0.873 0.884 102.764
29 CNN_DO_MP_DO_X3_64_128_256-1 0.735 0.718 0.716 0.759 0.837 0.857 72.193
30 CNN_DO_MP_DO_X3_64_128_256-2 0.750 0.745 0.739 0.713 0.776 0.789 101.459
31 CNN_DO_MP_DO_X3_64_128_256-DNN 0.716 0.703 0.703 0.811 0.850 0.866 114.463
32 CNN_DO_MP_DO_X3_64_128_256-Pat-3-0 0.731 0.720 0.731 0.773 0.834 0.840 104.831
33 CNN_DO_MP_DO_X3_64_128_256-Pat-3-1 0.724 0.714 0.711 0.789 0.872 0.883 68.565
34 CNN_DO_MP_DO_X3_64_128_256-Pat-3-2 0.761 0.759 0.749 0.683 0.766 0.788 107.973
35 CNN_DO_MP_DO_X3_64_128_256-Pat-2-0 0.748 0.741 0.732 0.724 0.795 0.819 87.255
36 CNN_DO_MP_DO_X3_64_128_256-Pat-2-1 0.750 0.741 0.735 0.716 0.785 0.802 91.753
37 CNN_DO_MP_DO_X3_64_128_256-Pat-2-2 0.719 0.703 0.702 0.806 0.899 0.914 62.646
38 CNN_DO_MP_DO_X3_64_128_256-Pat-1-0 0.676 0.688 0.688 0.929 0.994 1.005 41.329
39 CNN_DO_MP_DO_X3_64_128_256-Pat-1-1 0.673 0.682 0.686 0.933 0.998 1.012 42.944
40 CNN_DO_MP_DO_X3_64_128_256-Pat-1-2 0.674 0.698 0.698 0.938 1.010 1.021 50.324
41 CNN_DO_MP_DO_X3_64_128_256-Pat-4-0 0.757 0.751 0.747 0.698 0.764 0.781 115.440
42 CNN_DO_MP_DO_X3_64_128_256-Pat-4-1 0.777 0.767 0.768 0.638 0.706 0.705 162.051
43 CNN_DO_MP_DO_X3_64_128_256-Pat-4-2 0.737 0.696 0.687 0.748 0.890 0.915 83.437
44 CNN_DO_MP_DO_X3_64_128_256-Pat-5-0 0.752 0.740 0.722 0.707 0.787 0.814 100.885
45 CNN_DO_MP_DO_X3_64_128_256-Pat-5-1 0.780 0.756 0.756 0.636 0.717 0.734 142.063
46 CNN_DO_MP_DO_X3_64_128_256-Pat-5-2 0.739 0.722 0.720 0.743 0.831 0.848 83.833
In [ ]:
%%shell
jupyter nbconvert --to html /content/MSDS458_Assignment2_NB1_2.ipynb